

ChatGPT Advertising Banks Is Not One Story, It Is Two
ABN AMRO says its ChatGPT ads are working better than expected. A handful of earlier advertisers on the same channel say the opposite. Both are telling the truth about their own campaign.
By Katie Delaney / 2026-10-11 / 9 min read

Chatgpt advertising banks, the optimistic version
A fox does not judge a whole wood by one clearing. ABN AMRO's clearing, so far, looks good. The Dutch bank has run chatgpt advertising banks campaigns since 21 September 2026, and after two weeks it had plenty to say about them, reported first by Adformatie on 7 October 2026.
Ada Jong, ABN AMRO's head of digital sales and engagement, told the trade outlet that the mortgage campaign alone had already produced more than 1,000 concrete interactions, and that the bank's seven campaigns together had reached 1.5 million impressions in a few weeks. Her verdict, in her own words translated from Dutch: the click-through is better than expected.
Impressions across ABN AMRO's seven ChatGPT ad campaigns, reported 7 October 2026
ABN AMRO's own press statement went further, claiming engagement on business account openings scored many times higher through ChatGPT than through its social and display channels, with new customers, Dutch and English speaking alike, arriving through the ad specifically to open an account. Jong was careful, though, to call it a pilot, not a strategy, and flagged that the bank is still researching whether it can even deploy OpenAI's own measurement pixel internally, a caution any regulated institution will recognise.
Digital banking marketing ai is being treated as research, not a rollout
That caution matters. Digital banking marketing ai decisions inside a bank move through compliance and data-governance review long before a campaign goes live, and Jong's comments to Adformatie read like someone documenting an experiment, cross-checking ChatGPT performance against the bank's existing Google Ads dashboard rather than declaring victory and reallocating budget. OpenAI's own help centre, in Ads in ChatGPT: The Basics, updated 9 October 2026, describes the ad format, delivery and measurement advertisers like ABN AMRO are now working with directly.
The other clearing looks different
A month earlier, the same outlet reported a far less flattering picture. On 7 September 2026, Adformatie covered a handful of early international pilots that found low conversions, click counts that did not reconcile with Google Analytics, and tight EU targeting limits, since ChatGPT ads in Europe are not yet personalised to chat history and depend on conversation context alone.
| Item | Value |
|---|---|
| Choice OMG campaigns | 2 to 0.6 |
Choice OMG's managing director Peter Jaffray ran three small Ontario and Alberta campaigns for roughly $415, generating almost 9,000 impressions and 60 clicks, a 0.6 percent click-through rate against the roughly 2 percent his other channels usually return. His read, as reported: ChatGPT behaved more like a display channel than an engaged search audience, and none of the clicks converted into an interested buyer.
| Item | Value |
|---|---|
| OpenAI-reported clicks | 57 of 20 |
That gap is the one worth sitting with longest. Ai search ads fintech decisions lean on attribution the way a fox leans on scent, and if the platform's own count and the advertiser's own analytics cannot agree on a shared number, every other claim about performance inherits that same uncertainty.

Why both results can be honestly true
Chatgpt advertising banks results depend heavily on who is asking, what they are asking about, and how new the tooling was when they tried it. The September tests ran on an earlier version of ChatGPT Ads, before OpenAI added self-serve CPC bidding on 5 May 2026, and well before OpenAI's new visual ad format and expanded measurement tools, announced 5 October 2026, two days before ABN AMRO's upbeat report.
ChatGPT Ads will tell you the test is working. Then someone asks what actually hit the bank. If those three cannot be mapped to the same number, the report is a story, not a result.
Chatgpt ads financial services campaigns also differ by intent in a way a Canadian SaaS test or a B2B lead-generation funnel simply does not. Jong told Adformatie that Dutch users increasingly search ChatGPT with concrete mortgage questions: what this costs, whether it fits their finances, phrasing closer to a decision than a browse. A mortgage query carrying that much intent is a different advertising surface than a cold B2B SaaS pitch, even on the identical platform and the identical week.
Five lessons worth taking into a neobank advertising strategy
None of this says avoid the channel. It says treat it the way ABN AMRO is treating it: as a pilot with guardrails, not a budget line to copy from a headline.
Before trusting any reported performance, compare OpenAI's own numbers against your own analytics for the same window. A 57-versus-20 gap is not a rounding error.
A high-intent query, like a mortgage calculation, behaves differently on ChatGPT than a cold product pitch. Budget for the campaigns that mirror genuine decision-stage searches.
Context-only targeting, with no chat history, is a genuine constraint on precision. Creative has to work harder when targeting cannot.
ABN AMRO is still reviewing whether it can deploy OpenAI's measurement pixel. Start that review before the campaign, not after the results come in.
Internally and externally, call this what ABN AMRO calls it: an early experiment, not a settled channel. Neobank advertising strategy built on two weeks of data is a hypothesis, not a plan.
A bank's caution is the actual signal here
The most useful detail in the ABN AMRO story is not the 1.5 million impressions. It is Ada Jong's insistence on pairing every ChatGPT result against the bank's own Google Ads dashboard before drawing a conclusion. That discipline, applied consistently, is what keeps an enthusiastic early report from becoming an unearned budget decision six months later.
folkfox runs these reconciliation checks for fintech clients testing chatgpt advertising banks placements alongside paid search and paid social, so a two-week pilot produces a decision, not a press quote repeated past its evidence.

The wider prowl: everyone is chasing this trail
Chatgpt advertising banks is not the only AI ad surface moving fast through autumn 2026. At the same Advertising Week in New York where ABN AMRO's early figures were circulating in trade coverage, Meta and TikTok each announced their own agentic advertising features, reported on 9 October 2026 by Music Ally. Meta's business assistant is gaining memory, able to recall past campaigns and seasonal patterns; TikTok is pushing further into agentic buying tools of its own.
Read against each other, the pattern is a thicket of platforms all betting the same bet at once: that a conversational or agentic layer sitting on top of an ad auction will out-earn the old click-and-scroll model. A bank weighing chatgpt advertising banks spend against Meta or TikTok's new agentic tools is not choosing between a safe channel and an experimental one. It is choosing which experimental channel to run first, and every one of them is still thin on independently verified results.
That is the honest frame worth taking into a 2027 budget conversation. Digital banking marketing ai spend across any of these platforms deserves the same scrutiny ABN AMRO is applying to ChatGPT: pair every headline figure against your own measurement, segment by genuine query intent, and track the gap between what the platform reports and what your own systems can independently confirm. To outfox a crowded field of agentic ad launches means moving early but verifying constantly, not betting the whole den on the first promising number a vendor hands you.
A fox that moves first into a fresh clearing still checks the ground before it settles. The banks and fintechs doing the same with chatgpt advertising banks spend, testing small, documenting honestly, cross-checking every figure, are the ones whose neobank advertising strategy will still make sense once the platform's own novelty wears off and the real, reconciled numbers are all that is left on the table.
Frequently asked questions
How did ABN AMRO's ChatGPT advertising campaign perform?
After two weeks, ABN AMRO reported over 1,000 interactions from its mortgage campaign alone and 1.5 million impressions across seven campaigns, with click-through described as better than expected. The bank still calls it a pilot.
Have other companies had good results with ChatGPT Ads?
Results are mixed. Canadian agency Choice OMG recorded a 0.6 percent click-through rate against a usual 2 percent on other channels, and B2B agency Cleverly found OpenAI's reported click count did not match its own Google Analytics figures.
Why is there a measurement gap between OpenAI and Google Analytics?
The exact cause was not confirmed publicly, but the pattern, a higher platform-reported click count against a lower independently measured visit count, is the kind of attribution mismatch that makes any single performance claim on a new ad channel hard to take at face value.
Can banks target ChatGPT ads to specific chat history in the EU?
No. In the European Union, OpenAI currently targets ads using the context of the current conversation only, not stored chat history, which limits how precisely a financial advertiser can target a campaign.
Who sees ads inside ChatGPT?
Only users on the Free plan or the lower-cost Go plan. Plus, Pro, Business, Enterprise and Edu account holders do not see ads, according to OpenAI's own help documentation.
Should every bank try chatgpt advertising banks placements now?
Not without a measurement plan. Chatgpt advertising banks results so far range from a cautious early win at ABN AMRO to weak, hard-to-verify results elsewhere, so a reconciled pilot beats a blind rollout.
Ready to test ChatGPT Ads without guessing at the numbers?
folkfox runs AI search ad pilots for fintech clients with the measurement reconciliation built in from day one, not added after a headline needs checking.
Want folkfox in your Google results and AI answers? Set folkfox as a preferred source.

