The Real State of AI in Mortgage Marketing (From Someone Who's Been Doing This 30 Years)

I've been doing mortgage marketing since I was Employee #7 at DeepGreen Bank in 2000 (we grew to about 80 people running what felt like a billion-dollar bank). I built EquityOnline at Quicken Loans. I owned and ran Velocity Lending, a DTC mortgage lender, from 2016 to 2018 — my proof-of-concept that the playbook I'd been selling to Kaleidico clients actually worked when I ran it myself. I've run Kaleidico's mortgage practice since 2005 and today serve as its CRO. And through BRSG, I advise several leading home equity and mortgage fintechs on content and demand generation.

So I have a useful perspective on the "AI in mortgage marketing" conversation that's dominating the industry right now. The perspective is: most of what you're reading is hype, a meaningful subset is actually working, and the industry is going to split into two groups over the next 24 months based on who figures out the difference.

Let me lay out what I'm actually seeing.

The hype that's not working

A lot of mortgage lenders in 2025 and 2026 have rolled out AI-powered marketing initiatives that aren't really producing measurable lift. The common patterns:

"AI chatbots" that are just chatbots with a fresh coat of paint. The mortgage industry has been experimenting with chatbots on loan officer landing pages and lender websites for years. Most of them use generic LLMs with prompts that assume general-purpose queries. They give wrong rate information, can't handle state-specific regulatory questions, and hallucinate on product details. Conversion lift relative to a good contact form: essentially zero.

AI-written blog content at scale. Several mortgage lenders have commissioned hundreds or thousands of AI-generated articles in the last 18 months, usually as part of a "programmatic SEO" play. A small number of these programs are working (well-supervised, with human editorial discipline). Most are not — Google has rolled out multiple algorithmic updates specifically targeting low-effort AI content, and the sites with unsupervised AI content have watched traffic decline. The lenders running these programs often don't know how badly they're doing because they're tracking traffic in aggregate, not by cohort.

"AI-powered lead scoring." Every lead scoring platform in 2026 has the word "AI" somewhere in its product page. In practice, most of them are running logistic regressions or gradient boosted trees on historical lead data — techniques that have existed for 20 years. The "AI" branding doesn't change the underlying math. More importantly, the scoring only works if your lead data is clean and your attribution is correct, and neither is true at most lenders.

Marketing automation platforms with an "AI assist" button. You know the experience. The existing platform gets an "AI-powered email subject line generator" bolted on, and the vendor reprices it as an AI product. The outputs are mediocre because the model has no context about your brand, your regulatory environment, or your customers. Net lift: marginal at best.

AI-generated video content. The "AI avatar" tools have improved dramatically, but the application to mortgage marketing has been weak. Generic AI-generated loan officer videos don't build trust, they erode it. The mortgage consumer wants to know there's a real human accountable for the biggest financial transaction of their life.

If you're a mortgage marketer and any of this sounds familiar, you're not doing it wrong — most of the industry is in the same place. But you should know that the incremental revenue lift from these initiatives is mostly theater.

The AI that's actually working

Now the useful part. There are specific applications of AI in mortgage marketing right now that are producing real, measurable impact. I know because I'm either building or operating several of them.

Content production systems with strong human editorial oversight. This is the big one, and the nuance matters. An AI-assisted content production system, where an experienced editor is directing the AI, reviewing outputs, and enforcing quality standards, produces content at 3-5x the throughput of a pure human operation at 70-80% of the editorial quality. For topics where perfect editorial quality isn't required (most informational content), this is a huge economic advantage. For topics where perfect quality IS required (compliance-sensitive content, rate tables, specific regulatory guidance), pure human review is still necessary. The trick is knowing which is which. The lenders that are winning with AI content have invested in editorial systems that make that distinction automatically.

Compliance review assistance. LLMs are actually good at flagging compliance-sensitive language in marketing copy. Before a mortgage ad goes to the compliance team for formal review, running it through an AI-assisted compliance pre-check catches 60-80% of the issues. This doesn't replace compliance review — it speeds it up dramatically and reduces the number of rejected drafts.

Lead intake and routing augmentation. Using LLMs to interpret natural-language intake on lead forms and route leads to the right queue based on the customer's actual situation (refinance vs. purchase, state, loan size, urgency) works better than rule-based routing. The difference shows up in contact rates and conversion rates.

SEO content planning. Using AI to analyze SERP coverage, identify content gaps, and generate editorial briefs that a human writer then executes on is quietly one of the highest-ROI uses of AI in mortgage marketing right now. The gain isn't in the content generation — it's in the research and strategic planning layer.

Voice of customer analysis. Running LLMs across large volumes of call recordings, chat transcripts, and form submissions to surface customer themes, objection patterns, and language gaps is producing insights that human analysts would take months to reach. This informs messaging, script updates, and content strategy.

Personalized outbound at scale. Not mass email with a merge field — genuinely personalized outbound sequences where the AI drafts initial messages based on specific customer context (life event signals, search behavior, product fit) and a human loan officer reviews and sends. This works because the personalization quality is high enough to get responses that generic outreach doesn't.

The pattern across all of these: AI is an augmentation layer on top of good marketing fundamentals, not a replacement for them. The teams that are getting real lift from AI have strong marketers and loan officers who are using AI to do more of what they already do well. The teams that are getting no lift are trying to use AI to do things the team doesn't understand in the first place.

What's coming in the next 12-18 months

A few things I'm confident about, based on what I'm seeing in the companies I work with:

The AI answer layer will become a major source of mortgage lead origination. Consumers are increasingly starting their mortgage research inside ChatGPT, Claude, Perplexity, and Gemini. "What are the best HELOC lenders?" "How do I know if I should refinance?" The lenders whose content is consistently cited by these AIs will get the leads. The lenders whose content is not cited will see their organic funnel shrink, not because their site traffic dropped but because consumers never made it to their site in the first place.

Compliance-aware AI content systems will separate winners from losers. The mortgage industry has a compliance envelope that generic AI tools don't respect. The lenders that build (or buy) AI content systems with specific compliance guardrails — fair lending, TILA/RESPA, TCPA — will be able to move faster than competitors without regulatory risk. The lenders that try to bolt on generic AI will either move slowly (good compliance, slow output) or move fast into trouble (fast output, compliance problems).

Personalization will get genuinely good. The current state of marketing personalization is embarrassing — "Hi [FIRST NAME]" emails in 2026. The tooling is rapidly maturing to the point where mortgage marketing can genuinely personalize content to life events, credit profiles, and financial situations at scale. The lenders that build this capability will have a meaningful conversion advantage over competitors sending generic email blasts.

Loan officer-level AI assistants will be the sales floor investment of 2027. Every LO will have an AI assistant that knows the lender's products, the current rate sheet, compliance guidelines, and the specific customer's situation. The LOs who use these well will produce dramatically more funded loans per hour than the ones who don't. This will change LO compensation structures and probably headcount planning.

The industry will realize too late that AI-generated mortgage content has hurt its credibility. Consumers can often tell when they're reading slop, even if they can't articulate why. Mortgage lenders that flooded their blogs with AI content in 2025 are going to have trust rebuild work to do in 2027-2028. Some of them will consolidate and clean up. Some will just keep declining.

The test I'd apply to every AI initiative

If you're a mortgage marketer evaluating an AI project right now, here's the test I use:

What specific marketing outcome does this change, and by how much?

If you can't answer that question with a number attached to a timeframe, the project is probably hype, not substance.

"Use AI to write more content faster" is not an outcome. "Produce 100 high-quality pieces per quarter at 60% lower cost, resulting in 40% more organic traffic and 25% more organic leads within 12 months" is an outcome.

"Use AI for lead scoring" is not an outcome. "Improve contact-to-application rate from 12% to 15% by re-scoring the top of funnel" is an outcome.

The outcome-focused framing forces you to reckon with what the AI is actually going to change, and whether the change matters. Most AI mortgage marketing projects fail that test. The ones that pass are the ones worth doing.

The broader principle

Thirty years of fintech marketing experience has taught me that every technology wave in this industry — the internet itself, search marketing, social media, display advertising, marketing automation, mobile, each of them — followed the same pattern. A hype phase where everyone claimed to be using the new thing, a disillusionment phase where most of the initiatives quietly died, and a consolidation phase where a small group of operators who actually figured out the technology produced outsized results for the next decade.

AI is in the hype phase right now in mortgage marketing. We are 18-24 months from the consolidation phase. The lenders that use this period to build real capabilities, real systems, and real editorial discipline will be the ones who own the consolidation phase.

The ones who keep hiring "AI-powered" vendors, launching "AI chatbots" on their websites, and publishing AI-generated content without editorial review will spend the consolidation phase rebuilding what they broke.

If you're a mortgage marketing leader, don't get too excited about the current state of AI in your space. But do not ignore it either. The durable competitive advantage in mortgage marketing over the next decade is going to go to the teams that figured out, in 2025-2027, which AI applications actually work and which are theater.

I work on exactly these questions at Bill Rice Strategy Group and build AI-assisted content and marketing systems at Verified Vector. If you're navigating this in your mortgage operation, get in touch.