Getting Cited by AI When Compliance Demands Disclaimers

Open ChatGPT. Ask it whether you should take a personal loan at 1.2% flat monthly interest.

It will decline. It will explain that the decision depends on your income and existing obligations, which it does not know. It will suggest consulting a licensed financial advisor.

No citation. No brand named. Nothing an institution could have optimised for.

Now ask it how effective interest is calculated from a flat monthly rate.

It answers. It explains the difference between flat and declining balance. It cites sources.

Same product. Same user. Same session, possibly. And the entire strategy for financial GEO sits in the gap between those two responses.

The Engines Draw a Line You Did Not Draw

Generative engines treat financial questions differently from every other category, and they do it deliberately.

Ask about a moisturiser and you get a product recommendation. Ask about software and you get a comparison. Ask whether to take a specific loan and you get a refusal wrapped in a disclaimer.

The line runs between information and advice, and the engines enforce it because giving bad financial advice to a stranger has consequences that giving bad skincare advice does not.

Which sounds like a limitation until you notice who else cannot occupy it.

The Trust Paradox Nobody Is Exploiting

Four independent studies point the same direction, and the direction is strange.

EY surveyed 18,152 people across 23 countries and found 49% had used AI for savings or investment decisions in the previous six months. Half of consumers, consulting a machine about their money.

YouGov found that only 20% express any trust in AI within financial services. The lowest reading across every sector they measure. Lower than healthcare. Lower than legal.

TD Bank surveyed 2,504 people and found 78% use AI tools, while just 18% would trust one with a financial recommendation.

Northwestern Mutual found 13% trust AI for retirement planning, against 56% who trust a human advisor.

Read those together and the strategy writes itself.

Consumers ask AI about money because it is fast and free. Then they seek confirmation from a source that carries accountability, because the machine does not. That confirmation is the only thing a licensed institution can supply and a generative engine structurally cannot.

Not as a marketing claim. As a fact about who answers to a regulator.

Where Your Answer Sits on the Page Decides Whether It Is Read

Roughly 44% of citations that generative engines extract come from the first 30% of a page. The final third contributes less than a quarter.

Models behave like impatient editors. They read the opening, take what answers the question, and move on.

Now consider how financial content is typically written.

Nobody wrote it this way out of carelessness. The disclaimer went first because it felt safer, and the safety was real, and the cost was invisible.

The same content, restructured

Compliance does not require the disclaimer to appear first. It requires the disclaimer to appear.

Both versions satisfy every regulatory obligation. Only one of them gets cited.

What the Peer-Reviewed Research Actually Found

Most GEO advice is vendor marketing. There is one exception worth knowing.

Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan and Deshpande published the first peer-reviewed study of Generative Engine Optimization at ACM SIGKDD 2024. They built a benchmark called GEO-BENCH, tested ten thousand queries across nine domains, and measured what actually moved visibility.

Two findings from that paper matter enormously for financial institutions, and almost nobody quotes them.

The Law and Government domain benefited most from adding statistics. That domain is the closest available proxy for regulated financial content. Citing regulators and quantifying claims is not merely good compliance practice. It is the single most effective tactic available in this category, and it has been measured.

Lower-ranked sites gained more than already-dominant ones. A regional bank has more to gain from GEO than a national one. This inverts the usual dynamic of search, where authority compounds and incumbents pull away.

Financial institutions already produce verifiable facts, cite regulators, and publish statistics, because compliance demands it. The paper says that is exactly what generative engines reward. Most institutions have simply buried it beneath three paragraphs of disclaimer.

The Channel That Vanished in Six Weeks

Before building any GEO strategy in this category, understand what happened in September 2025.

Reddit's citation share inside ChatGPT collapsed. Investors noticed before marketers did, and Reddit's stock fell 14.4% over five trading days.

How far did it fall? Depends who measured.

The cause remains contested, which is the part most articles skip.

The popular explanation is that Google removed its num=100 search parameter in mid-September, cutting off the deeper search results that third-party data providers fed to OpenAI. Many Reddit threads rank below position twenty.

Semrush's own Head of Organic and AI Visibility disputes it. Of the 263 million keywords Reddit ranks for in Google, only about 90 million sit between positions 21 and 100. That is 34%, which does not account for a collapse this steep. His alternative: OpenAI deliberately reduced over-citation of specific domains, to be less biased and more resistant to manipulation.

Nobody outside OpenAI knows. What everybody knows is that a channel lost most of its value in six weeks because something changed in a configuration file, and no announcement was made.

Why this matters more in finance than anywhere

BrightEdge tracks citation volatility by category. Financial services is the most volatile they measure. More than half of financial domains see their citation position change week to week, and roughly nine in ten of those changes are declines.

An annual AI visibility audit in this category measures a position that stopped existing in week three.

Where Each Engine Goes Looking

A strategy built for one engine tells you almost nothing about the others, and financial queries expose the difference sharply.

That Indonesian anomaly deserves attention. A GEO strategy copied from a Western agency underweights Perplexity by a factor of seven, in a market where it reaches a fifth of AI users.

The Regulator Is Watching This Too

One more thing has changed, and it moves GEO from a marketing concern into a compliance one.

POJK 6/2026, issued in June 2026, governs financial influencer content. The critical detail is where the penalty lands. Administrative sanctions of up to fifteen billion rupiah fall on the licensed institution, not on the influencer it engaged.

OJK published an AI governance framework for the financial sector in 2025.

Read those together and the direction is unmistakable. What an AI says about your product is becoming your responsibility. Citation accuracy stops being a visibility question and becomes an audit question.

Which means off-site citation building in financial services, the practice of encouraging third parties to reference your institution, has to be conducted as a compliance exercise before it is conducted as a marketing one.

What Actually Gets Built

All of this rests on technical foundations that generative engines do not provide. AI crawlers rely on the same infrastructure search crawlers use. A financial site Google cannot index properly will not be cited well by ChatGPT either, however immaculate the structured data.

Fintech SEO is not an alternative to Generative Engine Optimization for financial services. It is the ground the second one stands on, and the authority signals that satisfy Google's quality raters are the same ones that make content citable.

I wrote Cited or Silent specifically about this problem, because the gap between institutions that understood the citation mechanics and institutions that were still writing longer disclaimers was widening every quarter, quietly, in a channel nobody was measuring. The companion volume, Found Before They Search, covers the search foundations everything else rests on. Both are free as gated editions, and both are published in paperback and hardcover and listed on Google Play Books and Apple Books.

What Silence Costs

Nothing that shows up anywhere.

A borrower somewhere asks an AI how effective interest works. The engine explains, cites an aggregator, and the borrower reads the aggregator's comparison table. Then they choose a lender.

That never appears in your analytics as a loss, because it never touched your site. No impression to compare against. No click-through rate declining. No flag anywhere.

Half of consumers now consult AI before financial decisions. Only a fifth trust what it tells them, and they go looking for someone accountable to confirm it.

You are either that someone or you are not.

Frequently Asked Questions

Why do AI engines refuse to answer some financial questions?

There is a line between information and advice, and the engines enforce it. Ask how effective interest is calculated and you get a cited answer. Ask whether you should take a specific loan and you get a refusal, a disclaimer, and a suggestion to consult a licensed advisor. Financial institutions can only occupy the first territory, and the entire GEO strategy for this sector follows from accepting that.

If consumers do not trust AI for financial advice, why does GEO matter?

Because they consult it anyway. EY found 49% of consumers used AI for savings and investment decisions in six months, across 18,152 respondents in 23 countries. YouGov found only 20% express any trust in AI within financial services, the lowest of any sector. TD Bank found 78% use AI tools while just 18% would trust it with a recommendation. People ask the machine, then look for a credible source to confirm what it said. Being that source is the opportunity.

How do we get cited when compliance requires disclaimers everywhere?

Move the disclaimer rather than removing it. Roughly 44% of citations are extracted from the first 30% of a page, so an answer buried under three paragraphs of qualification never gets read. Sentence one carries the answer, sentence two the regulatory context, sentence three the required disclaimer. Identical compliance posture, entirely different citation outcome.

What does the peer-reviewed research say about GEO?

Aggarwal and colleagues published the first peer-reviewed GEO study at ACM SIGKDD 2024, testing ten thousand queries across nine domains using a benchmark called GEO-BENCH. Citing sources, adding quotations and adding statistics improved visibility by 30 to 40 percent. Keyword stuffing produced zero effect and sometimes a negative one. Crucially, lower-ranked sites gained more than dominant ones, and the Law and Government domain, the closest proxy for regulated financial content, benefited most from statistics.

Did Reddit's citation share really collapse in September 2025?

Yes, and how far depends entirely on who measured. Spotlight tracked a fall from 14.29% to 0.21% of all cited sources. SEMrush tracked roughly 60% to 10% of prompt responses. An agency study reported to RBC Capital said 29.2% to 5.3%. PromptWatch measured 9.7% to 2% of ChatGPT answers. All four are correct and measure different quantities. The cause is contested: Semrush's own analyst disputes the popular explanation that Google's num=100 removal was responsible, noting only 34% of Reddit's rankings sit in positions 21 to 100.

How often should we monitor AI citations in financial services?

Weekly. BrightEdge tracks financial services as the most volatile citation category, with more than half of financial domains seeing their position change week to week, and roughly nine in ten of those changes being declines. An annual AI visibility audit measures a position that stopped existing in week three.

Does JavaScript-injected schema work for AI crawlers?

No. AI crawlers read the server-side HTML head before JavaScript executes. Schema injected after page load renders correctly in a browser and validates in testing tools, but the crawler finds nothing. On a financial site this means rates, terms and licence numbers are legally displayed to a human and completely invisible to every engine deciding what to cite.

Why does Perplexity matter more in Indonesia?

Telkomsel bundled Perplexity Pro into prepaid and postpaid data plans from 28 May 2025, the first connectivity and AI bundle in Indonesia. Perplexity now holds roughly 11 to 21 percent of Indonesian AI chatbot share against a global average nearer 2 to 3 percent. A strategy copied from a Western agency underweights it by a factor of seven.

Are we liable for what an AI says about our products?

The regulatory direction suggests increasing institutional responsibility. POJK 6/2026 places sanctions for financial influencer content on the licensed institution rather than the creator, with penalties reaching fifteen billion rupiah. OJK published an AI governance framework for the financial sector in 2025. Citation accuracy is becoming an audit question rather than only a visibility one.

---