Demand engines in mortgage lending
The evidence base for how volume is actually produced in this category, the compliance controls required to produce it safely — and the live register of what we are building, in what order, and what it has returned so far.
Activity record
What has been done
A running log of work performed on this engagement, most recent first. Every finding in the Research tab traces back to an entry here.
Organic social audit — client Instagram and competitive benchmark
Client Instagram profile and most recent post audited; competitive set benchmarked on the same day for audience and content volume. Three competitor handles could not be verified and are excluded rather than reported as zero.
Output: Social tab. Key finding: 104 followers and 11 lifetime posts, dormant since April, against a business holding 2,802 five-star reviews — a content supply problem, not an audience problem.
Competitive advertising intelligence — full inventory capture
Every active advertisement across the competitive set captured and coded: launch date, run duration, variant duplication, offer, format, call-to-action and account structure. 1,300+ active advertisements reviewed across six advertisers.
Output: Competitor ads tab. Key finding: LendingTree has run one home-equity creative continuously for 443 days while every rival's oldest live creative is under five months — harvesting versus still testing. Bankrate near-tripled its inventory in four days.
AI-search measurement assessment
Established what is actually knowable about consumer AI-assistant use in lending, and separated measured data from vendor modelling. Reviewed the tooling market (Semrush AI Toolkit, Ahrefs Brand Radar, Profound, Peec, Otterly, Evertune) on methodology and price.
Output: Research tab, Part Three. Key finding: prompt-level volume does not exist from any credible source; first-party referral tracking is free and back-dateable twelve months.
Regulatory verification pass
Ten areas checked against primary sources: telephone-consumer consent and revocation, state telemarketing statutes, trigger leads, platform ad-targeting restrictions, Regulation Z advertising, Regulation N, RESPA Section 8, FTC review rules, data safeguards, and fair-lending posture.
Output: Research tab, Part Two — eight operating controls. Three prior assumptions corrected, including the status of one-to-one consent and which rule governs home-equity advertising.
Paid-creative inventory across the competitive set
Public ad libraries pulled for all six operators; creative, longevity and duplication recorded; offers and terms catalogued into a product inventory.
Output: Part One §1.4 and §1.5. Key finding: the category's top creative has converged on one offer — equity access that preserves an existing first-mortgage rate.
Competitive teardown — six operators
Consumer funnels walked step by step and recorded; public filings read for scale, marketing spend and channel mix; search footprints sampled. No personal information was submitted and no forms completed.
Output: Part One §1.1–1.3 and §1.6, five charts. Key finding: five structurally different business models, only one of which supports heavy paid media.
Action register
What we are building, in order
Twenty-one actions across four phases. Sequencing is deliberate: measurement first because it back-dates, controls second because they block spend, demand engine third because it compounds, and paid media last because it only pays once the rest exists.
Phase 0 — Measurement foundation · this week · no cost
Create a custom channel group matching assistant referrers, positioned above the referral rule. Because channel groups apply at query time, this reclassifies historical sessions — producing a twelve-month month-over-month trend immediately, at no cost. This is the single highest-value action on the list and the answer to the question that prompted Part Three.
Define and instrument the funnel: sessions, quote starts, completed leads, contact within five minutes, applications, funded loans — by channel and product. Without this, nothing later in the register can be evaluated.
Weekly grep for user-triggered fetches — the signal that an assistant retrieved a page because a human just asked something, as distinct from training crawls. Begin archiving now; log retention is typically 30–90 days and cannot be recovered later.
Twenty-five to forty real buyer prompts run monthly across the major assistants, with answer text archived verbatim and timestamped. At this scale a disciplined manual panel rivals a paid tool — and the archive doubles as an advertising-substantiation record under Control 6.
Phase 1 — Compliance controls · before any acquisition spend · blocking
Written to the strictest state in the operating footprint, not the federal floor. Federal one-to-one consent was vacated in 2025, which makes state statutes — several with private rights of action and five-figure per-violation damages — the binding constraint.
One list propagating to dialer, texting platform, email and ad audiences within one business day, capturing verbal opt-outs logged by loan officers alongside keyword replies. Ten-business-day honoring has been law since April 2025.
Every vendor contract carries an express warranty of statutory compliance and indemnity, and produces the consent artifact itself rather than an attestation that one exists.
Declare before the classifier does it mid-flight. Written targeting standards prohibiting audience exclusions and proxy targeting. Platform policy requires fresh confirmation — the court supervision that produced the current rules expired in June 2026.
One gate covering paid social, SMS, landing pages, rate tables and loan-officer-authored posts. Must specifically cover the home-equity negative-trigger rule — the most likely thing a lending marketing team gets wrong given the category's convergence on equity offers.
Twenty-four-month retention floor with substantiation attached, capturing creative at publication rather than reconstructing it later. Note that distributing templated creative across individual loan-officer pages multiplies this obligation rather than escaping it.
Every builder, realtor, advisor or platform arrangement documented with a market-value basis and the services actually rendered, before it starts. Criminal exposure and a private right of action attach here.
Ask every closed customer through the same flow, disclose incentives and employee relationships, never suppress or selectively display. Separately, inventory CRM, dialer and ad platforms as in-scope for safeguards and state privacy law, with opt-out and browser-signal handling verified end to end.
Phase 2 — Demand engine · weeks 3–10 · compounding
Annual equity reviews and rate-watch alerts against the closed-client book. Evidence: the largest retail lender in the study sources most refinance production this way at a 73% recapture rate. Highest return per dollar in the register.
A guarantee operations can actually meet, with a customer-side deadline that makes it deliverable. Distinct naming is now essential — the generic five-day equity convention is heavily contested. Claims and naming route through Control 5 before launch.
Local rate and programme pages carrying the freshness and named-reviewer mechanics the category's publisher uses, plus product and question content. Concede national head terms. Note the finance-specific finding: two-thirds of AI-Overview citations in this vertical come from sources not ranking in the top hundred — citability and ranking are separate games.
Individual originator pages carrying their own verified reviews. The one asset in this study no national operator has built — the largest has no named humans in its consumer experience at all.
Sub-five-minute contact in business hours, with routing and alerting to a named person. A shared marketplace lead typically reaches three to five lenders simultaneously; response time decides the outcome.
Real-estate teams, builders, advisors, credit unions and employers, treated as a measured channel rather than referral hope. Gated on Control 7. Evidence: the fastest-growing lender in the study grew funded volume 89% on flat advertising by moving distribution here.
Phase 3 — Amplification · month 3 onward · only once the above holds
Local and product-specific only. Paid media at scale requires resale economics this business does not have — the marketplace in the study recovers a 73%-of-revenue marketing ratio by selling each consumer up to five times.
Only with sub-minute dialed response, tight filters, and a hard funded-loan cost ceiling agreed in advance. Default assumption is that the same budget buys more as owned demand.
Segmented sequences for purchase, refinance, equity, first-time and not-yet-ready, with consent language cleared under Control 1 before anything sends.
Monthly reporting in funded volume by channel and product, not clicks. Closes the loop opened by Baseline 0.2.
Measurement
Baseline — awaiting instrumentation
The metrics this engagement will be judged on. Every row reads pending because nothing is instrumented yet; Phase 0 changes that, and the AI-referral row back-fills twelve months on day one.
| Metric | Source | Baseline | Reporting cadence |
|---|---|---|---|
| AI-assistant referral sessions | Analytics channel group | Pending — 12 months back-fills on setup | Monthly |
| User-triggered AI page fetches | Server logs | Pending — no archive before setup | Weekly |
| Share of AI answers citing the brand | Manual prompt panel | Pending — cannot be back-dated | Monthly |
| Sessions → quote starts | Analytics | Pending | Monthly |
| Quote starts → completed leads | Analytics | Pending | Monthly |
| Leads contacted within 5 minutes | CRM + call system | Pending | Weekly |
| Leads → applications | CRM | Pending | Monthly |
| Cost per funded loan, by channel | CRM + finance | Pending | Monthly |
| Recapture rate on closed-client book | CRM + servicing data | Pending | Quarterly |
Dependencies
What we need to begin
- Analytics and web-property access, plus server-log access or an export path.
- An introduction to compliance counsel, and confirmation of the licensed-state footprint — Control 1 is written to the strictest state, so the list determines the standard.
- CRM and loan-origination system access, so leads route into tools the team already uses and funded-loan attribution can close.
- Loan-officer roster with individual licence identifiers, for the named-human pages in 2.4.
- A decision on the flagship offer for 2.2, and whether the closed-client book in 2.1 is available for marketing use under existing consents.
- Ad-platform account access, to confirm category declaration status under 1.4.
Part One
How the volume is actually produced
Every company in this study is described in the market as a "lead generator." That label conceals five distinct business models with radically different cost structures. Sorting them is the precondition for deciding what to imitate.
1.1The five mechanics
Ranked by how reliably each produces volume for an operator that originates its own loans, rather than by how much revenue it currently produces for its owner.
Re-engaging customers already in the book — through servicing relationships, credit monitoring, or rate-watch alerts — produces transactions at near-zero incremental media cost. It is the only mechanic here whose unit economics improve with age, and the only one a competitor cannot outbid you for.
Placing origination capability inside someone else's existing audience — a personal-finance platform, a homebuilder, an advisor network — substitutes a revenue share for customer-acquisition cost. Volume arrives pre-qualified by the host's relationship with the consumer.
Ranking on the queries a borrower types immediately before transacting — rates, product mechanics, local terms, comparisons — and converting on the page. The mechanic that compounds: the asset appreciates while the media bill does not recur.
Naming a process guarantee, attaching a customer-side deadline that makes the promise operationally deliverable, and letting trade and local press carry it. The product doubles as the campaign.
Buying traffic in volume and monetizing it faster than it costs. This works when one acquired consumer can be resold several times; it fails when the acquirer must fund the loan to realize revenue. Most lenders should treat it as a targeted supplement, not a growth engine.
1.2The economics of demand
Three charts explain most of the strategic divergence in this category. All figures are from public company reporting for the periods labeled.
LendingTree FY2025; Better and loanDepot Q1 2026. Different period lengths, same ratio.
The red bar is the arbitrage business. LendingTree can spend at this ratio because it resells each consumer; Better and loanDepot must fund loans to earn, and their ratios reflect that constraint. A lender that benchmarks against the red bar is copying an economic model it does not have.
Share of funded volume by channel. Total volume grew 89% over the same period.
The single most important strategic chart in this study. Better did not out-advertise anyone into 89% growth — it changed where volume comes from. Management targets 60% platform share by year-end 2026.
Total revenue $1,117.3M. Mortgage is the smallest and least-favored line.
Strategically important for anyone considering buying LendingTree mortgage leads: mortgage is 14% of their business, and their Home segment profit fell 24% year over year in Q1 2026 on higher media costs. Their attention and their best media inventory follow insurance.
1.3Company teardowns
Six operators, ordered by model. Financial figures are from public reporting for the periods stated. Lead prices are industry-reported estimates and labeled as such.
| Model | Not a lender. Registered as a marketing lead generator and licensed broker. Matches a consumer request to multiple lenders and charges an upfront match fee for each transmission — its filings state "up to five match fees may be generated from a single consumer loan request form." Paid whether or not a loan closes. |
| FY2025 scale | Revenue $1,117.3M. Insurance $711.9M · Consumer $253.4M · Home $151.8M. Roughly 770 network partners; homepage claims 300+ lenders, "149M people helped," "$297B in loan funding." |
| Marketing | $812.9M selling and marketing (~73% of revenue); ~69% of revenue is variable media, reallocated in real time toward whichever vertical clears best. |
| Funnel design | Roughly eighteen sequential micro-questions with pre-filled sliders and auto-advance. Timeline intent is asked first ("Already under contract" through "Not sure"), and a real-estate agent question mid-funnel harvests a second sellable lead. Contact fields appear only at ~86% completion, flanked by encryption and no-credit-impact reassurance. |
| Consent infrastructure | Third-party consent witnessing plus device fingerprinting and risk tokens on every step — the consent artifact is effectively the product being manufactured. |
| Retention asset | Spring, a free credit-monitoring product: 31.3M cumulative users, ~0.9M initiating revenue-generating transactions in 2025. |
| Search footprint | ~1,772 curated URLs — deliberately small. Ranks #1 for "home equity loan rates"; concedes daily rate-news queries. |
| Lead economics | Estimated $30–100 per shared mortgage lead, $80–150+ exclusive (industry-reported). Non-exclusive by design; typically 3–5 simultaneous buyers, making speed-to-dial decisive. |
| Sentiment | Trustpilot 4.5 of 5 across ~17,000 reviews. Dominant negative theme is contact volume after submission; a fresh telephone-consumer action was filed during the research window over messages sent after an unsubscribe. |
| Model | On-page rate marketplace inside an editorial property. The consumer self-selects a lender and clicks out; Bankrate earns a fixed fee on connection. Its own disclosure: "we earn a fixed fee if you connect with the lender." |
| Conversion surface | Personalization inputs render a lender table showing rate, APR, monthly payment, points, upfront costs and an eight-year total cost — sorted by total cost, not headline rate. Two clicks from landing to monetized handoff. |
| Homepage | Rebuilt as a brand funnel: "9 out of 10 homebuyers overpay for their mortgage. You don't have to." Trust stack cites $73k average saved, 850 banks surveyed, founded 1982, 100M+ users. |
| Freshness system | Rates stamped to the morning of the visit ("Rates as of Friday, July 31, 2026 at 6:30 AM"); "Best mortgage lenders of August 2026" published July 31 — the month-ahead title rolled a day early. |
| Authority system | Separate written-by, edited-by and reviewed-by credits naming a licensed senior mortgage loan originator, an "Expert verified" badge, published methodology ("surveyed 75 lenders"), and five separate editorial-standards pages. |
| Local scale | Fifty-one state rate pages carrying genuine local content — state assistance programs, median sale prices, insurance-market commentary. |
| Search results | Top-five for "current mortgage rates" and #3 for "HELOC rates" via rate-table URLs; #7 for "mortgage calculator"; only #9–10 on "best mortgage lenders," where large media brands dominate. |
| Lead economics | Estimated $200–250 per mortgage connection (industry-reported) — several times LendingTree's price, for a consumer who chose the lender rather than being sold to one. |
| Model | Digital direct lender that also licenses its Tinman origination platform to partners, earning volume through their customer bases. |
| Q1 2026 scale | Funded volume $1.64B (+89% YoY), split evenly direct-to-consumer ($824M) and platform ($821M); platform grew 404% YoY. Revenue $47.5M (+52%); net loss $70.3M. Product mix: refinance 52%, purchase 36%, HELOC 12%. |
| Marketing | $9.2M in the quarter versus $8.7M a year earlier — a 6% increase supporting an 89% volume increase. |
| Partners | Intuit's Credit Karma (140M members, under 1% penetrated), NEO Home Loans (loan officers saw 91% growth in funded loans six months post-implementation; run rate $1.50B to $2.97B), Finance of America, Coinbase. A prior Ally Bank channel disappeared when Ally exited mortgage origination in 2025 — the cautionary case for partner concentration. |
| Margin signal | Gain-on-sale ~2.5 points direct-to-consumer, ~3.5 points through NEO, and 6–7 points on HELOC — the clearest product-margin signal in the study. |
| Conversion framing | Primary CTA carries microcopy on the button itself: "3 min | No credit impact." Funnel opens with an AI assistant offering purchase, refinance or HELOC before requesting any personal detail. |
| Automation | "Betsy" voice AI: ~100,000 calls per month in 2025, 1.89M cumulative, 35.5% of borrower inquiries resolved with no human involvement, ~1,666 loan-officer hours saved monthly. |
| Paid social | No active advertising inventory found — including for its own flagship branded products. |
| Model | Largest originator by unit volume, running a direct channel plus the Rocket Pro wholesale channel serving broker partners — West Capital Lending among them. Ecosystem includes Redfin, Rocket Money and Rocket Loans behind a single mandatory account. |
| Trust stack | "Over 9 million mortgages closed"; J.D. Power servicing awards cited 2002–2025; "America's largest lender," footnoted to 2025 HMDA origination units. |
| Funnel | Single-CTA discipline — effectively every page routes to "See what I qualify for." Account creation is mandatory before real numbers appear, converting the funnel into an ecosystem entry point. Speed framing throughout: "10 minutes or less," "no credit impact." |
| Content | ~2,600 educational URLs with a full parallel Spanish library, author pages and editorial-standards infrastructure, plus live timestamped rate tables for all fifty states. |
| Rate presentation | Live table timestamped to the hour. But advertised rates assume a 740 credit score and carry 1.625–2 discount points — the 30-year fixed headline rate carries 1.75 points, about $6,125 upfront on a $350,000 loan. Disclosed, and materially different from what a shopper believes they are comparing. |
| Offers | ONE+ (1% down, Rocket contributes 2%); $12,000 back through the Redfin bundle; VA 0% down. |
| Paid social | Largest active inventory in the study, ~360 ads — and the impression leaders are all home-equity creatives. |
| Exploitable gap | No named individual anywhere in the consumer experience; testimonials are first-name only; service runs through a call center and chat. |
| Model | Consumer-direct retail lender on the proprietary mello platform, with homebuilder joint ventures and a 2025 re-entry into wholesale. Founder Anthony Hsieh returned as CEO in July 2025. |
| FY2025 scale | $26.5B originated (+8%), revenue $1.19B (+12%), net loss $108M — a fourth consecutive loss year; ~1.4% market share. Q1 2026: $7.66B originated, only 41% purchase. |
| The engine | A $120.7B servicing book across 455,634 loans converting at a 73% organic refinance recapture rate (up from 71%). Rate-locks grew 14% in a quarter while marketing spend fell 12%. The company reinvested $60M into mortgage servicing rights — buying future lead flow. |
| Marketing | $29M in Q1 2026, down from $33M the prior quarter — a deliberate retreat from the 2021 era of stadium naming rights and national television. |
| Equity strategy | The equityFREEDOM suite (first- and second-lien HELOCs plus fixed home-equity loans), extended April 2026 through a Figure partnership embedding third-party underwriting into mello — the "5x5 HomeLoan," five-minute approval and five-day funding, with a projected $1B+ in new HELOCs. |
| Retention device | "Lifetime Guarantee" — waived fees on future refinances, a recapture mechanism dressed as a benefit. |
| Sentiment | Zillow 4.88 of 5 across ~4,700 reviews, with praise concentrated on individual loan officers. Complaints skew to servicing (196 CFPB mortgage complaints in 2024, mostly payment-process issues). BBB A+. |
| Structural weakness | No rates published on the marketing site; a thin non-brand content footprint (~55 learning-center URLs, absent from generic cash-out search results); and documented loan-officer attrition — its October 2025 suit alleges a competitor recruited 175+ originators and diverted customers using stolen data. Allegations only; contested, with a counterclaim pending. |
| Model | Wholesale brokerage grown through loan-officer recruiting and acquisitions into Rocket Mortgage's #1 national broker partner. Co-founded by Eric Hines and Daniel Iskander. |
| Claimed scale | $9B produced in 2025 across 1,300+ originators in 44 states; $20B+ lifetime. Company-reported figures. |
| Primary channel | Industrial outbound augmented by voice AI. One documented deployment: 318,000+ outbound calls producing ~11,400 live conversations and 1,788 warm handoffs to 70 originators, with a reported 180% increase in booked appointments. Job listings confirm a consumer-direct pod working "LIVE leads provided daily." |
| Paid social | ~250 active ads distributed across thirty-plus individual loan-officer pages using centrally templated creative — reach multiplication without central page spend. CTAs favor Messenger and text-to-lead ("text HELOC to…"). |
| Signature offer | A 5-Day HELOC: no appraisal via automated valuation, no tax returns or pay stubs, soft pull only, up to 85% CLTV and $750K, 620 minimum FICO. Secondary offers cover DSCR, bank-statement and hard-money payoff. |
| Review asset | ~40,384 reviews at a 4.89 average via automated post-closing collection, linked directly from advertising — social proof operated as an acquisition asset. |
| The cost | A public complaint record dominated by contact practices: messages continuing after a stop request (one account describes four further texts within a minute, from different originators), calls from rotating numbers, calls to do-not-call registered numbers, and a review alleging "quotes with rates they don't have." BBB A+ but 52 complaints in three years. |
1.4Product and offer inventory
What each operator actually sells, and the specific offers carrying the weight. The convergence is visible here before it appears in the creative: five of six lead with home equity.
| Company | Named offers and products | Terms as advertised | Lead product in paid media |
|---|---|---|---|
| Rocket Mortgage | ONE+ · Redfin bundle · home equity loan · VA | 1% down (Rocket adds 2%); $12,000 back; rates at 740 FICO + 1.625–2 pts | Home equity — "keep your current mortgage rate," AVM instead of appraisal |
| loanDepot | equityFREEDOM · 5x5 HomeLoan · Lifetime Guarantee · mello | 5-min approval / 5-day funding; HELOC + fixed HELoan, 10/20/30-yr; fee waiver on future refis | HELOC and HELoan — "no refinance required"; 580-FICO accessibility hook |
| West Capital | 5-Day HELOC · DSCR · bank-statement | No appraisal (AVM), soft pull, ≤85% CLTV, ≤$750K, 620 FICO | 5-Day HELOC across 30+ loan-officer pages |
| Better | One Day Mortgage · One Day HELOC · Better Forever · Betsy AI · Tinman | Commitment letter in one day (docs within 4 hrs of lock); HELOC decision 24 hrs, cash 7 days; $0 lender fees | None — no active paid social; PR and partners instead |
| LendingTree | Marketplace match · Spring credit monitoring | Up to 5 lender offers; soft pull to start; free to consumer | Auto insurance and credit cards lead; mortgage and home equity secondary |
| Bankrate | Rate tables · calculators · lender reviews | Ranked by 8-year total cost; "beat 99.7% of banks"; no hard credit check to see rates | Mortgage rate comparison — "no sales calls, just the numbers" |
Read the third column against the second. Three companies now market a five-something-day equity product, and two market a one-day mortgage decision. Speed in home equity is table stakes as of 2026, not a differentiator.
1.5Paid footprint and creative convergence
Public ad libraries disclose creative and duration for commercial advertisers, but not spend, impressions or conversion. Absent metrics, the standard practitioner proxies are longevity and duplication: an ad still running many months after launch, or replicated across many variants, is being scaled because it works.
Public ad-library counts, August 1, 2026. Commitment to the channel, not performance.
The two fastest-growing operators by percentage — Better and loanDepot — sit at the bottom. Paid social is where this category spends when it lacks a distribution or database advantage. Separately, 110 active ads use the exact phrase "5-day HELOC" across many advertisers — the offer is proven and the name is commoditized.
Industry-reported ranges, not published rate cards. Directional.
A shared lead typically reaches three to five lenders at once; a Bankrate connection reaches one, by consumer choice. The price gap is not quality inflation — it is exclusivity and intent. Every lead an operator's own pages capture is a $200+ connection it did not have to buy.
Finding: the category has converged on one offer
Among the operators running meaningful paid social, the highest-exposure mortgage creative is materially identical — access home equity without disturbing an existing low first-mortgage rate. Rocket's leading units read "Access your home's equity and keep your current mortgage rate… Calculate online with no credit hit," with automated valuation replacing an appraisal. loanDepot's inventory is predominantly equityFREEDOM carrying the same "no refinance required" framing. West Capital's entire program is the 5-Day HELOC. More than a hundred active ads use that exact phrase across many advertisers.
This is a rational response to a real condition: a large installed base of homeowners holding mortgages priced well below current rates, for whom a cash-out refinance is irrational but an equity draw is not. It also means the offer no longer differentiates. Advantage has moved to proof and execution.
Finding: soft entry outperforms hard application prompts
LendingTree's longest-running mortgage creative — active without material change since September 2025 — pairs a loss-framed message ("It's never been more important to compare mortgage rates to ensure you're not overpaying") with an interactive home-price selection visual and a "Calculate Your Payment" call to action. Rocket's equity ads open with quiz-style questions ("How soon do you need funds?"). Bankrate's entire campaign is a creator-style talking-head video carrying one auditable statistic.
- Tool before form. "Calculate" and "compare" consistently outrank "apply" in surviving creative.
- Risk removal at the click. Better prints "3 min | No credit impact" on the button itself; Rocket uses "no credit hit" in ad copy. Time cost and credit fear are handled before the landing page.
- Verifiable claims beat adjectives. Bankrate's campaign rests entirely on "we tracked 300+ mortgage lenders weekly; our top rate beat 99.7% of them."
- Contact restraint is now a selling point. Bankrate advertises "no sales calls, just the numbers" — direct evidence that the category's dominant negative experience has become a competitive opening.
1.6What transfers to a smaller balance sheet
Not every mechanic is available to an operator without a servicing book, a platform to license, or nine-figure media budgets. Sorted by transferability:
| Mechanic | Transferable? | What the smaller version looks like |
|---|---|---|
| Database recapture | Fully | Closed-client equity reviews and rate-watch alerts replicate loanDepot's 73% effect without owning servicing. Highest return per dollar in this table. |
| Commercial-intent search | Fully, at local scope | Concede national head terms — Bankrate itself only reaches #9 on "best mortgage lenders." Own local rate and program pages with its freshness and named-reviewer mechanics. |
| Partner distribution | Fully, at relationship scale | Real-estate teams, builders, advisors, credit unions and employers replace Credit Karma. Same economics: revenue share instead of media cost. |
| Branded speed product | Fully | Requires a guarantee operations can meet and a customer-side deadline that makes it deliverable — Better's four-hour document rule is the model. Distinct naming now essential. |
| Named-human proof | Uniquely available to smaller operators | Rocket has no named humans at all. Rate.com has thousands of loan-officer pages but no per-officer reviews. Nobody in this set pairs named originators with their own verified reviews. |
| Paid media at scale | No | Requires LendingTree's resale economics. Viable only as targeted local and product-specific spend on top of a converting funnel. |
| Industrial outbound | Technically — see Part Two | West Capital's mechanic works. The complaint and litigation exposure it generates is the subject of the next section. |
Part Two
Operating guardrails for a regulated brand
Every mechanic in Part One is legal to run. Several are also the fastest routes to a consent order. This section converts the applicable rules into eight operating controls, each with the date its status was verified — because three of the rules most often cited in mortgage marketing materials changed in the last eighteen months, and one is now the opposite of what most guidance still says.
2.1Read the enforcement shift correctly
The most consequential development for a lending marketer in 2025–2026 is not a new rule. It is a redistribution of who enforces the existing ones.
Federal posture has genuinely softened. The Bureau dismissed a large share of pending enforcement actions and is targeting a major reduction in enforcement headcount; it withdrew roughly seventy guidance documents in May 2025; and in a rule effective July 21, 2026 it removed the effects test from Regulation B and stated that the Equal Credit Opportunity Act does not authorize disparate-impact liability. HUD withdrew its algorithmic-advertising guidance in September 2025 and has proposed rescinding its disparate-impact rules.
It would be expensive to read that as reduced exposure. The statutes are unchanged, the Supreme Court precedent recognizing disparate-impact claims under the Fair Housing Act still stands, and a regulation disclaiming a statutory theory is itself vulnerable to challenge. Meanwhile the vacuum is filling: state attorneys general brought multiple settlement-services actions in the first quarter of 2026, a fair-lending matter settled for $68 million in March 2026, California's privacy regulator is actively fining marketing-side data practices, and at least fifteen states now run telephone-consumer statutes stricter than federal law with private rights of action attached.
2.2The eight controls
Each control states the rule, the operating requirement, and the verification date. Citations are provided so counsel can go directly to source.
What changed: the FCC's one-to-one consent rule — the requirement that a consumer consent to each seller individually, widely described as closing the lead-generator loophole — was vacated by the Eleventh Circuit in Insurance Marketing Coalition v. FCC on January 24, 2025, and the text was formally deleted from the Code of Federal Regulations effective August 29, 2025. The federal standard reverted to 47 C.F.R. § 64.1200(f): a signed written agreement clearly authorizing the seller to deliver telemarketing by automated means. Multi-seller "you and our marketing partners" consent is again federally permissible.
That is precisely why the federal standard is the wrong build target. At least fifteen states impose stricter requirements with private rights of action: Florida, Oklahoma, Washington and Maryland set the pattern; Texas SB 140 took effect September 1, 2025 with statutory damages up to $5,000 per violation plus treble damages and fees; Oregon followed January 1, 2026; Pennsylvania carries an October 18, 2026 compliance deadline.
Under 47 C.F.R. § 64.1200(a)(10), effective April 11, 2025, a consumer may revoke consent by any reasonable method. Stop, quit, end, revoke, opt out, cancel and unsubscribe are per se reasonable, as is a verbal request to any employee. Revocation must be honored within ten business days. The broader provision extending one revocation across unrelated message categories is deferred to January 31, 2027.
Two 2026 decisions appear to narrow exposure — one circuit holding the do-not-call provision does not reach text messages, one district court holding consented texts fall outside quiet-hours limits. Both are contradicted elsewhere and neither is settled. Several states mandate an 8pm cutoff regardless.
The Homebuyers Privacy Protection Act, signed September 5, 2025 and effective March 2026, amended FCRA § 604(c) (15 U.S.C. § 1681b(c)). A credit bureau may furnish a mortgage trigger lead only where it is a firm offer of credit and the recipient is one of four categories: holder of the consumer's documented opt-in consent, originator of the consumer's current mortgage, current servicer, or an insured depository or credit union holding an existing account for that consumer.
Mortgage advertising falls into restricted categories on the major platforms — Meta's Financial Products and Services category (which replaced the former Credit category in October 2024) and Google's Consumer Finance category, which names home loans explicitly. Operationally: no ZIP-code targeting, no age or gender targeting, a fifteen-mile minimum radius on Meta and one kilometre on Google, no lookalike or similar-audience expansion, and significant customer-list restrictions. Meta classifies advertisers automatically from creative, so an undeclared campaign gets categorized anyway — usually mid-flight.
These restrictions originate in a 2022 Department of Justice settlement whose court supervision expired June 27, 2026. Whether Meta retains its variance-reduction system and current category rules absent the decree is unverified — treat 2027 planning as requiring fresh confirmation. More durably: the fair-lending statutes are unchanged, private plaintiffs and state regulators retain full authority, California's automated-decision regulations took effect October 1, 2025 with four-year recordkeeping, and Texas's AI statute took effect January 1, 2026.
Regulation Z § 1026.24(d)(1) defines four triggering terms for closed-end credit, and the list is exhaustive: amount or percentage of down payment, number of payments or period of repayment, amount of any payment, and amount of any finance charge. State any one and § 1026.24(d)(2) requires the corresponding disclosures — down payment, repayment terms reflecting the full term including any balloon, and the annual percentage rate using that exact phrase, plus a statement if it can increase.
Two points commonly missed. First, home-equity lines are governed by § 1026.16(d), not the closed-end rule, and its triggers fire on terms stated negatively as well as positively — "no closing costs" and "no annual fee" are triggering terms. Second, § 1026.24(i) separately prohibits seven practices outright, including misleading comparisons to hypothetical rates, misleading use of a consumer's current lender's name, and foreign-language advertising that states triggering terms in the foreign language while making disclosures only in English.
The Mortgage Acts and Practices rule, Regulation N, 12 C.F.R. § 1014.3, prohibits material misrepresentation — express or implied — across nineteen enumerated categories covering rates, fees, taxes and insurance, prepayment penalties, variability, product type, government affiliation, the source of the communication, and the ability to refinance. It binds not only lenders and brokers but advertising agencies, lead generators and rate aggregators.
Section 1014.5 requires retention for twenty-four months from last dissemination of every materially different commercial communication, plus sales scripts, training materials and marketing materials, and documentation of the products actually available. Enforcement runs through the Bureau, the FTC and state attorneys general.
RESPA § 8 prohibits giving or accepting anything of value for the referral of settlement-service business, and prohibits unearned fee splits. Marketing services agreements remain lawful only where payment is for services actually performed at market value and is not tied to referrals; co-marketing requires each party to pay its proportionate share of actual market value.
Two pieces of guidance survived the 2025 purge and both bear directly on Part One's partner and lead-buying mechanics: the 2020 RESPA § 8 FAQs, and the February 2023 advisory opinion on digital mortgage comparison-shopping platforms, which holds that non-neutral presentation that steers consumers, combined with payment tied to that steering, is a prohibited referral fee. Federal enforcement has receded; state enforcement has not, and § 8 carries criminal exposure and a private right of action.
The FTC's Rule on Consumer Reviews and Testimonials, 16 C.F.R. Part 465, effective October 21, 2024, carries civil penalties up to $53,088 per knowing violation. It prohibits fake reviews, purchased reviews, reviews procured from employees or their immediate relatives, compensation conditioned on a particular sentiment, undisclosed insider reviews, company-controlled sites presented as independent (§ 465.6), and review suppression (§ 465.7) — including displaying a filtered subset while implying it represents all reviews.
On review gating — soliciting only customers expected to respond favorably — the honest answer is that FTC staff guidance states the rule contains no specific prohibition, but that the practice could still violate Section 5 of the FTC Act. Not per se banned; also not safe. This matters directly to any operator copying West Capital's 40,000-review automated collection flywheel.
On data: the Safeguards Rule, 16 C.F.R. § 314.4, expressly covers mortgage brokers as well as lenders, requiring a named qualified individual, written risk assessment, encryption in transit and at rest, multifactor authentication, annual penetration testing, service-provider oversight, a written incident-response plan and an annual board report — plus notice to the FTC within thirty days of a breach affecting 500 or more consumers, on a public docket. And the point most often missed: the financial-data privacy exemption covers information, not entities. Prospect and lead data collected before an application, advertising pixels, cookie identifiers and purchased lists sit fully inside state privacy law — where recent fines have landed, including one against a marketing firm for building custom audiences from consumer profiles.
2.3Failure modes observed in this study
These are not hypotheticals. Each was visible in the public record of an operator examined in Part One, and each maps to a control above.
- Contact after opt-out — West Capital. Its public complaint record is dominated by messages continuing after a stop request; one account describes four further texts within a minute of the request, sent from different originators. Under Control 2 this is a ten-business-day obligation with a private right of action in several states.
- Rotating caller identity — West Capital. The same record describes calls from continually changing numbers, including to do-not-call registered numbers. Rotation to evade filtering converts a compliance problem into an intent problem.
- Consent architecture is not immunity — LendingTree. Its funnel captures litigation-grade artifacts, including third-party consent witnessing and device fingerprinting at every step — and it still faces recurring telephone-consumer litigation, with a fresh action filed during the research window over messages sent after an unsubscribe. If the best-instrumented operator in the category is still exposed, an operator without that instrumentation is not managing the risk; it is unaware of it.
- Quoted rates that were not available — West Capital. A published review alleging quotes at rates the broker did not have implicates both Regulation Z § 1026.24(a) — advertise only terms actually available — and Regulation N's prohibition on misrepresenting rates.
- Disclosed but decision-changing pricing — Rocket. Advertised rates assume a 740 credit score and carry up to two discount points, roughly $6,125 upfront on a $350,000 loan. This is lawfully disclosed. It is also the clearest demonstration in the study that technical compliance and consumer clarity are different standards — and that the gap between them is a positioning opportunity for a competitor willing to close it.
- Recruiting as a data-transfer risk — loanDepot and West Capital. The pending suit alleging 175+ originators departed with customer data illustrates that aggressive loan-officer recruiting carries trade-secret and privacy exposure alongside the commercial upside. Allegations only; contested.
2.4Pre-launch checklist
These controls cost little to build before a campaign runs and a great deal to retrofit afterward. Minimum set before any acquisition spend begins:
- Consent language reviewed and approved by counsel — for the strictest state in the footprint, not the federal floor.
- Suppression list operational across dialer, texting platform, email and ad audiences, one-business-day propagation, verbal opt-outs captured.
- Lead-vendor warranties executed covering trigger-lead eligibility, consent provenance and indemnity.
- Ad accounts declared in the correct restricted category, with written targeting standards prohibiting exclusions and proxies.
- Claims-review gate live for every channel, with the home-equity negative-trigger rule specifically covered.
- Automatic creative and script archiving at a twenty-four-month retention floor, substantiation attached.
- Every partner arrangement papered with a documented market-value basis.
- Review solicitation universal and un-gated, incentives and relationships disclosed.
- Marketing systems inventoried as in-scope for the Safeguards Rule and state privacy law, with opt-out and browser-signal handling verified end to end.
Method and limitations
Research conducted July 28 – August 3, 2026. Consumer funnels were walked directly in a browser and recorded step by step; no personal information was submitted and no forms were completed. Financial figures are taken from public company reporting for the periods stated. Advertising inventory counts are public ad-library results on August 1, 2026; those libraries disclose creative and duration for commercial advertisers but not spend, impressions or conversion, so all performance language in Part One is explicitly a proxy based on ad longevity and variant duplication. Search positions were sampled through a search API and are directional rather than literal rankings. Lead and connection prices are industry-reported estimates, not published rate cards. Scale figures attributed to West Capital Lending are company-reported and not independently audited.
Part Three distinguishes throughout between measured, panel-derived and modelled data. Survey findings carry their sample size where published and are marked directional where it is not. Two figures sourced from indexed excerpts rather thansource documents are flagged inline for re-verification before external publication. Litigation described is pending and consists of allegations that have not been adjudicated.
Regulatory positions in Part Two were verified against primary sources on August 2, 2026; items that are unsettled or expiring are flagged inline with their dates. This document is commercial research and an implementation plan, not legal advice. Part Two should be reviewed and adapted by qualified mortgage-compliance counsel before implementation.
Part Three
AI search — what is measurable and what is not
Consumers are beginning to ask assistants which lender to use. This section establishes what can be evidenced about that shift, what cannot be evidenced by anyone at any price, and which of the two the market's tooling is actually selling.
3.1The direct answer on volume
- No first-party disclosure. OpenAI, Anthropic, Microsoft and Google publish no query- or prompt-level volume. The most granular public release is an academic paper sampling roughly 1.5 million conversations, which discloses aggregate scale — 700 million weekly users, 2.63 billion daily messages as of mid-2025 — against a three-bucket taxonomy with no finance, lending or provider-selection category and no query strings.
- The occupational index is not consumer demand. Anthropic's economic reporting classifies usage by occupational task, not consumer query, and is dominated by software work. No lending breakout.
- Every commercial "prompt volume" product is modelled. The largest vendor states in its own documentation that individual prompts are "often too specific and unique to measure directly," which is why its volume metric is published at topic level, derived from clickstream panels plus machine learning — and concedes that no platform can provide exact visibility numbers, only directional signals.
Operational consequence: a vendor quoting "X monthly searches for mortgage broker in ChatGPT" is selling a model, not a measurement. That number should not appear in a board pack or a client deck without the word estimated attached to it.
3.2What is evidenced: consumer behaviour in lending
Survey data is the strongest evidence base available, because it measures stated behaviour directly rather than inferring it from traffic. The mortgage-specific findings are the most decision-relevant material in this report.
| Finding | Source | Confidence |
|---|---|---|
| 76% of homebuyers would be comfortable letting AI shop mortgage lenders on their behalf. 89% would share personal financial details with a lender's AI tool; 68% trust AI-provided mortgage information; 32% already use AI somewhere in the homebuying journey. | Veterans United, June 2026 | Directional — sample size and methodology not published |
| 55% used AI for financial advice, up from 10% a year earlier. Gen Z 77%, Millennials 72%, Gen X 49%, Boomers 30%. | TD Bank, March 2026, n≈2,500 | Good — sample published |
| Among 18–29s, 26% used AI for financial or mortgage advice versus 17% who used a licensed broker or advisor. | Canadian consumer survey, July 2026, n=1,910 | Good — sample published; non-US market |
| 19% of US adults used AI for financial advice; 67% acted on the suggestion. | Wells Fargo, via industry press, April 2026 | Moderate — sample not published |
Same survey instrument, twelve months apart. TD Bank, n≈2,500.
A 5.5× increase in twelve months on a consistent instrument. Whatever the absolute referral numbers say, the behavioural shift underneath them is not marginal.
3.3What is evidenced: referral traffic
Panel and aggregate-analytics measurement of clicks arriving from assistants. Fast growth on a small base — both halves of that sentence matter.
- Financial services AI-sourced traffic grew 266% year over year in the 2025 holiday season, with roughly 105–158% year-over-year growth reported through 2026 (aggregate analytics across a large customer base; the 2026 figures were read from indexed excerpts rather than the source document and should be re-verified before external publication).
- AI referrals grew ~357% year over year but represent roughly 1% of web traffic — 1.13 billion AI referral visits against 191 billion from conventional search in a single month of 2025; an independent benchmark put AI referrals at 1.08% of total site traffic in late 2025.
- AI-referred traffic now converts 42% better than non-AI traffic — a reversal from 38% worse twelve months earlier. Low volume, unusually high intent. This is the number that justifies attention despite the small base.
- Assistant share is fragmenting. ChatGPT fell from roughly 76% to 53% of category web traffic in a year as Gemini rose to 27–28% and Claude to about 9% — a multi-assistant measurement plan is now required, not a ChatGPT-only one.
3.4AI Overviews in finance — the counterintuitive part
Finance is a suppressed category for AI Overviews, which is good news, with one structural caveat that is better news.
- Only about 7.8% of finance keywords trigger an AI Overview, against roughly 15.7% across all categories — Google appears to suppress them on real-time and financially sensitive queries.
- But when they do appear, ranking and citation are different games. In finance, only ~11.3% of AI Overview citations overlap with the top-ten organic results, and about two-thirds of cited sources do not rank in the top hundred at all. A smaller lender can be cited without out-ranking the incumbents.
- Click-through impact is genuinely contested. Published estimates of organic CTR decline on AI-Overview queries range from −15.5% to −61%, with a partial 2026 rebound in at least one longitudinal study, and one analysis finding zero-click rates actually fell. Do not present a single number as settled.
3.5The three measurement types
Most confusion in this category comes from conflating three different things sold under one label.
| Type | What it actually measures | Trustworthy for | Cost |
|---|---|---|---|
| First-party referral | Real humans who clicked through from an assistant to your site. Ground truth — but blind to answers that resolve without a click. | Business impact. The only type reportable as fact. | Free |
| Synthetic prompt monitoring ~90% of the tooling market |
Whether models name your brand when you ask a question list you wrote. Non-deterministic; the API model is often not the consumer model; no personalisation or chat history. | Competitive diagnostics and catching factual errors about your brand. Never market share. | $29–400/mo |
| Panel-inferred demand | What real people are estimated to be asking, modelled from opt-in clickstream panels and reweighted. Topic-level, not prompt-level. | Deciding which topics to create content for. | $99–400/mo |
3.6Why crawler traffic is not a usable proxy
A common shortcut is to read AI bot activity in server logs as a proxy for assistant usage. The platform data disproves it: published crawl-to-referral ratios run to roughly 50,000:1 for one major AI company and 887:1 for another. Crawling scales with model training, not with consumer queries, and around 80% of AI bot traffic has been measured as training-purpose.
The exception worth tracking is the narrow class of user-triggered fetches — an assistant retrieving a page because a human just asked something. That is a genuine demand signal and is distinguishable in logs, which is why it appears separately in the action register rather than being lumped with crawler volume.
3.7How this engagement will gather the data
Four collection methods, each with its access requirement and its honest limitation stated up front.
| Method | Mechanism | Requires | Cadence | Limitation |
|---|---|---|---|---|
| Referral analytics | Custom channel group matching assistant referrers, above the referral rule | Analytics access | Monthly | Misses answers with no click; app browsers strip referrers into Direct |
| Server-log fetches | Weekly grep for user-triggered agent strings, verified by reverse DNS | Log access | Weekly | Agent strings are spoofable; retention typically 30–90 days |
| Brand-visibility panel | Fixed set of 25–40 buyer prompts run across assistants; answer text archived verbatim with timestamps | Assistant accounts | Monthly | Measures model output, not consumer demand; results vary run to run |
| Competitive and regulatory watch | Re-pull of public ad libraries, competitor offer pages, and the flagged regulatory items | None — public sources | Monthly / on trigger | Public data only; vendor pricing and policy pages change without notice |
Two regulatory items carry known expiry dates and are on the watch list by default: the deferred cross-category revocation provision due January 2027, and platform advertising policy following the June 2026 expiry of the court supervision that produced the current targeting restrictions.
Competitive advertising
Paid creative intelligence
A structured read of every active advertisement the competitive set is running, captured August 5, 2026. Longevity, duplication, offer, format and call-to-action are recorded for each advertiser — because those are the variables the platform actually discloses.
4.1Inventory and longevity
Two measurements, and the relationship between them is the whole story. Volume shows commitment. Longevity shows conviction — an advertisement still running many months after launch is one the advertiser has decided works.
Captured August 5, 2026. Page-scoped counts except where noted.
Red bars use distributed models — the advertisement sits on an individual loan officer's page rather than the corporate one. Bankrate moved from roughly 54 active advertisements on August 1 to 150 on August 5, a near-tripling inside four days; that is an active scale-up, not a steady state.
Days between observed launch date and August 5, 2026. Longevity is the disclosed proxy for performance.
The single most useful number on this page. LendingTree has a home-equity creative that has run continuously since May 19, 2025 — over fourteen months, replicated across six variants. Nobody sustains a paid creative for fourteen months unless it clears. Every other advertiser's oldest live creative is under five months, which means they are still testing while LendingTree is harvesting.
4.2The comparison matrix
| Advertiser | Active ads | Oldest live creative | Max variants | Dominant offer | Format | Primary CTA | Structure |
|---|---|---|---|---|---|---|---|
| LendingTree | 430 | 443 days | 7 | Home equity comparison (evergreen); auto insurance (volume) | Mostly static | Get Quote | Central page |
| Rocket | 370 | 107 days | 3 | Equity without disturbing your rate; ONE+ 1% down | Video-led | Get Quote | Central page |
| West Capital | 220 | 146 days | 1 | 5-Day HELOC; DSCR and self-employed | Static | Send Message / text-to-lead | 26 officer pages |
| Bankrate | 150 | 63 days | 4 | High-yield savings; mortgage rate comparison | Static → video | Learn More | Central page |
| loanDepot | 130 | 50 days | 5 | HELOC / HELoan; refinance; apply-from-your-couch | Mixed | Learn More | Central + officer pages |
| Better | 0 | — | — | None — partner distribution and earned media instead | — | — | — |
Counts for West Capital and loanDepot are keyword-scoped because their inventory is distributed across individual loan-officer pages; the others are page-scoped. The two methods are not perfectly comparable and the distinction is preserved rather than smoothed over.
4.3The spend question, answered honestly
Per-advertisement spend is not disclosed. Total marketing spend is — for the public companies — and pairing actual disclosed spend against observed inventory is more useful than any per-ad estimate would be.
| Advertiser | Disclosed marketing spend | Period | Active ads | What the pairing shows |
|---|---|---|---|---|
| LendingTree | $812.9M | FY2025, selling & marketing | 430 | ~73% of revenue. Paid media is the business model; the inventory reflects it. |
| loanDepot | $29M | Q1 2026, marketing & advertising | 130 | Down from $33M the prior quarter — deliberate retrenchment, visible as a thin, very recent ad set. |
| Better | $9.2M | Q1 2026, marketing & advertising | 0 | Flat spend, 89% volume growth, zero paid social. The spend goes elsewhere entirely. |
| Rocket | Not separately disclosed at this granularity | — | 370 | National brand budgets; inventory is the only observable signal. |
| Bankrate | Private — within Red Ventures | — | 150 | Not disclosed. The four-day tripling is the observable event. |
| West Capital | Private | — | 220 | Not disclosed. Distributed structure spreads cost across officer budgets. |
4.4The creative that is actually working
Three advertisements, selected by the disclosed performance proxies — longevity and variant duplication — rather than by preference.
Running since May 2025 · six variants · static · "Get Quote"
Loss framing ("ensure you're not overpaying"), an interactive price-selection visual, and a calculation invitation rather than an application prompt. The home-equity sibling of this creative is the one that has run fourteen months. Note what is absent: no rate, no urgency, no discount.
Running since April 2026 · video-led · "Get Quote"
"Access your home's equity and keep your current mortgage rate… Calculate online with no credit hit." Roughly four in five of Rocket's active advertisements carry this offer, opened with quiz-style prompts ("How soon do you need funds?") and automated valuation in place of an appraisal.
Running since June 2026 · creator-style video · "Learn More"
One auditable statistic ("we tracked 300+ mortgage lenders weekly; our top rate beat 99.7% of them") delivered in creator format, with a sibling variant promising "no sales calls, just the numbers." The only advertiser in the set marketing the absence of contact — a direct read on the category's dominant complaint.
4.5The creative, verbatim
Actual copy from active advertisements, transcribed August 5, 2026. This is what each advertiser is literally saying — the patterns in §4.6 are drawn from these lines, not from impression.
| Core equity line (runs across most of the set) | "Access your home's equity and keep your current mortgage rate by getting a home equity loan. No hassle. No expensive fees. Calculate online with no credit hit." |
| Interactive openers | "How soon do you need funds? Two weeks" · "How much cash do you need? $50,000" · "Tap desired cash amount: $100,000" |
| Appraisal removal | "Aside from the fact that they don't affect your current mortgage, you can now get a home equity loan much faster with Rocket Mortgage's new automated valuation system, saving you time and money on appraisal fees." |
| Zero-cost framing | "Rocket Mortgage® is helping homeowners find the cash they need. Run the numbers for $0. What could unlocking cash from your home look like?" |
| Market-timing angle | "Take advantage of record high equity levels and see how much cash you could get from your home's equity." · "Home equity loans surge in popularity." |
| ONE+ down-payment offer | "You only need a 1% down payment to buy a home thanks to this program from Rocket Mortgage®." Shown with the arithmetic: "1% from you + 2% from us on a $350,000 home: $3,500 from you + $7,000 from us = $10,500 total down payment." |
| Affordability | "See how much home you can afford and estimate your monthly payment!" |
| The 443-day evergreen | "Unlock the value of your home with a home equity comparison" → "up to 5 offers in mins!" Running since May 19, 2025, six variants. |
| Data-led variant | "Americans are sitting on $28.7 trillion of home equity. Tap into yours!" |
| Use-case variant | "Use your home equity to pay for renovations. Compare multiple lenders now!" · "Unlock your home equity to pay for home renovations, repairs, and more!" |
| Mortgage (purchase) | "Are you shopping for a new home? It's never been more important to compare mortgage rates to ensure you're not overpaying for your loan." → "Calculate Your Payment" |
| Rate-news hook | "Lowest Mortgage Rates in 3 Years" · "Don't settle for the first lender you find! Comparing rates could save you tens of thousands over the life of your loan." |
| Non-mortgage volume | Auto insurance ("Rev Up Your Savings!" · "No Age Limits on Savings!"), health insurance, and business loans ("Compare up to 5 business loans without a hard credit pull"). |
| Convenience hook | "Apply from the convenience of your couch—no meetings or trips to the bank required. Get a no-hassle online rate quote in minutes." Five variants, refreshed Aug 3. |
| Equity, no-refi framing | "Your home has supported you, now let it support your goals. With loanDepot, access your equity through a fixed HELoan or..." · "Still need cash for a remodel, debt payoff, or unexpected bills? A HELOC or HELoan from loanDepot could help." |
| Debt-consolidation angle | "High-interest debt making upgrades feel impossible? You're not alone. Explore your options with loanDepot." |
| Refinance | "If rates have changed since you bought your home, refinancing could help you lower your mortgage rate and reduce your monthly payment." |
| Retargeting tell | "You've already started exploring your options, now's the time to take the next step" — explicit mid-funnel retargeting copy. |
| The claim | "We tracked 300+ mortgage lenders weekly. Bankrate's top rate beat 99.7% of them. Yours probably wasn't the exception." → "See what your bank can't match." |
| Variant | "Bankrate's best mortgage rates beat 99.7% of banks. Compare lenders in minutes and find the rate you actually deserve." |
| Anti-harassment | "A lower rate could mean huge savings on your mortgage. Compare rates from top participating lenders — no sales calls, just the numbers. 🔍 See offers in minutes 🔒 No hard credit check ✅ Free to use, no commitment" |
| Savings vertical | "Compare top high-yield savings accounts, find the fit for your specific goals, and open the account." · "Side-by-Side Comparison of Top Savings Accounts & Terms from Trusted Banks." |
| The 5-Day HELOC | "This 5 Day HELOC leverages technology to determine your eligibility within minutes! See your offer in 5 minutes" · "Turn Your Equity Into Cash, Fast! Check Your Rate For Free! No Underwriting. No Appraisal. No drawn out process." |
| Rate-protection framing | "Protect that great interest rate all while taking advantage of your home equity." |
| Debt consolidation | "Fast, efficient, hassle-free financing. Use a digital HELOC to borrow what you need to consolidate high interest debt." |
| Investor / self-employed | "💸 Self-Employed? Let Your Rentals Work for You. No income docs. No W-2s. Your property's cash flow is all you need." |
| Note | Identical copy appears across multiple named officer pages — centrally templated creative, locally distributed. |
4.6Five patterns worth copying
- The offer has converged; the execution has not. Equity access that preserves an existing first-mortgage rate is the dominant offer at Rocket, loanDepot and West Capital, and the longest-running creative at LendingTree. It is proven and it is crowded — advantage now sits in proof and delivery, not in the offer itself.
- Invite a calculation, not an application. "Calculate," "compare" and "run the numbers" appear in the surviving creative across every advertiser. "Apply now" appears mostly in advertisements less than sixty days old — the ones still being tested.
- Remove risk in the button, not the landing page. "No credit hit," "no hard credit pull," "for $0" and "3 min" sit inside the advertisement itself. The objection is answered before the click is paid for.
- A verifiable number outperforms an adjective. Bankrate's entire campaign rests on one auditable statistic; LendingTree's newest equity creative leads with "$28.7 trillion of home equity." Both are checkable, which is what makes them defensible as well as effective.
- Distribution multiplies reach cheaply — and multiplies obligation. West Capital's 220 advertisements sit across 26 officer pages from one templated creative. It works. It also means twenty-six surfaces to archive under the retention rule in Control 6, and twenty-six chances for a non-compliant claim.
Social channels
Organic social — audit and benchmark
Channel-by-channel audit of the client's own organic presence, benchmarked against the competitive set. Captured August 5, 2026. Additional channels will be added to this tab as they are supplied.
5.1Client channel audit
| Followers | 104 — against a business with 2,802 verified five-star reviews |
| Posts published | 11 lifetime |
| Following | 5 |
| Last post | April 14, 2026 — approximately 16 weeks before capture. No publishing activity since. |
| Engagement on last post | Zero comments. Like counts are not visible to logged-out viewers, so reach cannot be assessed from outside the account. |
| Bio | "At Lumin Lending, we guide you every step of the way with honest advice and tailored loan solutions." Carries NMLS 2716106 and DRE 02265104 — licensing disclosure present and correct. |
| Caption, last post | "At Lumin Lending, we guide you every step of the way with honest advice and tailored loan solutions. No confusion. No pressure. Just results." |
What is working
- Licensing disclosure is correct. NMLS and DRE identifiers in the bio, where they belong. Many competitors' officer-level accounts omit this.
- Voice is consistent with the brand position. "Honest advice… no confusion, no pressure" matches the anti-sales positioning and is genuinely differentiated in a category advertising rates and speed.
- Profile links to the site, so any traffic the channel does generate is measurable once the analytics work in Phase 0 is complete.
Channel inventory — to be completed
| Channel | Handle | Audience | Content volume | Last activity | Status |
|---|---|---|---|---|---|
| @luminlending | 104 | 11 posts | Apr 14, 2026 | Audited — dormant | |
| Awaiting | — | — | — | Pending supply | |
| Awaiting | — | — | — | Pending supply | |
| YouTube | Awaiting | — | — | — | Pending supply |
| TikTok | Awaiting | — | — | — | Pending supply |
| Google Business Profile | Awaiting | — | — | — | Pending supply |
5.2Competitive benchmark
Organic Instagram presence across the competitive set, captured the same day. Note that scale here belongs to the media and marketplace businesses — the lenders are markedly weaker, which is the opening.
Captured August 5, 2026, from public profile data.
Three handles could not be verified and are excluded rather than reported as zero: the rocketmortgage handle resolves to an account with a single post and placeholder naming, so it is almost certainly not Rocket's active presence; westcapitallending returns an empty parked handle despite the company's heavy paid activity; and Better's handle did not resolve. Absence of a verified handle is not evidence of absence from the platform.
5.3Recommended approach
Proportionate to the channel's real commercial value — which is meaningful but secondary to the funnel and search work already in the register.
- Convert the review corpus into the content engine. Thousands of five-star reviews naming individual officers is a renewable content supply that costs nothing to produce and is impossible for a national competitor to copy. This alone resolves the eleven-post problem.
- Publish the people, not the products. Every advertiser in the competitive set is selling a rate, a speed or a comparison, and not one is putting a named human in front of it. Social is the cheapest place to test that positioning before it reaches paid.
- Treat the channel as proof, not reach. A prospect who has been referred, or who has just read 2,800 reviews, will check the profile. It needs to look alive and licensed — it currently looks abandoned, which actively undercuts the trust the reviews built.
- Keep licensing in every profile and any officer-level account. The bio already does this correctly; the obligation extends to any distributed officer accounts if that model is ever adopted, and those surfaces fall inside the twenty-four-month retention rule in Control 6.
- Instrument before scaling. The profile links to the site, so once Phase 0 is complete, social's contribution becomes measurable rather than assumed — and the decision to invest further can be made on evidence.
Additional channels supplied by the client will be audited into §5.1 on the same basis: audience, content volume, recency, engagement where observable, licensing disclosure, and measurable link to the site.