Fintech AI-Visibility Scorecard

How AI assistants describe the leading fintech tools when a founder asks for recommendations — scored on Visibility, Citation, and Recommendation. Every score below is derived only from verbatim, page-cited transcripts, so any reader can re-derive it.

Version 1 — Claude surface only. This is a point-in-time snapshot, not an audit. One of four intended AI surfaces is scored here (Claude). ChatGPT, Perplexity, and Gemini are pending founder-provided access and are shown empty — never estimated or simulated.

Why this matters for a fintech

We cannot tell you how many of your buyers ask an AI assistant before they ask anyone else, and nor can anyone else. What you can find out is what the answer says when they do.

Category answers skip your domain

0 of 4 engines cited our own domain when asked a “best tool” question. They cite third-party listicles instead.

One answer is not the answer

up to 60% of the sources an engine cites change between two identical questions asked the same day.

Source: CitePack’s own A/A probe ledger, 2026-07-18 — 137 grounded category probes and 12 same-prompt pairs per engine, measured against citepack.com itself. We publish our own numbers rather than borrowed vendor statistics. If an AI assistant does not name, cite, or recommend your product when a founder asks for the category, you are absent from the shortlist before a human ever compares features. This scorecard shows exactly where each fintech stands on that surface today.

How it was measured

Reproducibility header — the exact conditions this column was scored under.

Engine

Claude, via the model API

Captured

10 July 2026, 14:04 IST

Browsing

Off, no live retrieval

Access

Model called directly, not the web app

Model version

We did not record it, so this run cannot be reproduced exactly.

The scorecard — Claude surface

surface_score = Visibility + Citation + Recommendation (0–6). One tool per fintech category; each scored strictly against the browsing-OFF condition above.

#ToolVis.CiteRec./6
1Stripestripe.com2226
2QuickBooksquickbooks.intuit.com2226
3Unitunit.co2226
4Rampramp.com2226
5Mintretired / redirecting1102
6Mercurymercury.com2226
7Stripe Radarstripe.com/radar2226
8Affirmaffirm.com2226
9Bill (Bill.com)bill.com2226
10Coinbasecoinbase.com2226
Visibility — named unprompted among leading options (0–2)Citation — named its own canonical domain from knowledge (0–2)Recommendation — actively recommended (0–2)

Read the clustering honestly

Nine of ten tools score 6/6 — but that reflects three real conditions of this single surface at this moment, not a finished verdict: (a) browsing is OFF, so Citation collapses to “does the model know the canonical domain” — which it does for these well-known vendors, inflating Citation versus a browsing-on run; (b) each tool is the leader of its own category, so unprompted Visibility is naturally high; (c) with no live retrieval, the model leans on established incumbents. Scores will spread out once the browsing-on surfaces (ChatGPT / Perplexity / Gemini) are run, where citing a live own-domain link is a much harder bar. Treat this Claude column as one point-in-time input.

Composite view — 3 surfaces pending

The finished scorecard composites four surfaces (max 24 = 4 × 6). Three are intentionally empty until real access is provided — never estimated.

SurfaceStatusWhat it needsContribution
ClaudeScoredCaptured 2026-07-10, browsing OFFup to 6 / 6
ChatGPTPendingFounder’s logged-in chat.openai.com session or an OpenAI API key
PerplexityPendingFounder’s logged-in perplexity.ai session or a Perplexity API key
GeminiPendingFounder’s logged-in gemini.google.com session or a Google AI API key

Where your fintech would sit

This scorecard covers ten category leaders. To see your own product scored the same way — the exact question asked, the verbatim answer, and a per-line Visibility / Citation / Recommendation score you can re-derive yourself — CitePack runs the same glass-box probe on your category. Every number stays page-cited to a transcript; nothing is estimated. That transcript is the artifact you take into a positioning or GEO decision.

Claims-lint statement

This scorecard reports verified, point-in-time, page-cited observations of what the Claude surface said at the recorded timestamp. It is not an audit; it is not “complete”; it makes no guarantee and no “100%” claim. It is one of four intended surfaces — three remain pending — so it is a partial, point-in-time snapshot by design.

Glass-box source — every score is re-derivable from the verbatim probe transcripts (CitePack task T-0115).

Run it on your own product

The free scan reads your page, tells you which of the seven on-page signals need work, and asks Perplexity one question about your category. Asking all four engines whether they name you, cite you and recommend you is the $49 report. No signup either way.

See if AI cites your site.

Paste a URL. We ask a real engine what your customers ask, show you the answer it gave and who it named, and — if something is stopping it reading you — the line in your own file. Nothing is held back.