How to track AI citations for clients
There's no ranking API for AI answers, so you measure visibility by sampling: freeze a set of buyer questions, run them through ChatGPT, Perplexity and AI Overviews logged-out each month, and log whether the brand is cited and which sources the engine trusts. Citation rate is the headline metric — and the "who gets cited instead" column is your content and outreach backlog.
Why you sample, not query
Google rank tracking works because there's a results page with positions you can read. AI answers don't have that — there's no endpoint that returns "you rank #3 in ChatGPT." Each answer is generated, sources vary by phrasing and session, and personalization shifts what you see. So the only honest way to measure AI-search visibility is to sample it the way a buyer would experience it: ask the questions, logged-out, and record what comes back.
That makes method consistency the whole game. The same prompts, the same engines, the same logged-out conditions, the same monthly cadence — change any of them and you're comparing noise to noise. Freeze the instrument and the trend line becomes real. (This is also why GEO is sold as citation readiness, never a guaranteed placement — the engine still decides; see does GEO replace SEO?)
Six steps, repeated monthly
- 1
Freeze a prompt-set of the questions your client’s buyers ask
Write 15–20 questions per client — category ("best running shoes for flat feet"), buyer-intent ("X vs Y, which is better"), and brand ("is [brand] any good"). Freeze the exact wording: this set is your measurement instrument, and changing it breaks month-over-month comparability. The questions are the ones real buyers type into ChatGPT, not keywords.
- 2
Run each prompt across the AI engines, logged out
Run every prompt through ChatGPT, Perplexity, Google AI Mode / AI Overviews, Gemini and Copilot in a fresh, logged-out session so personalization and history don’t skew the result. Record the engine and the date. There is no "ranking API" for AI answers — visibility is measured by sampling, so a consistent method is what makes the data trustworthy.
- 3
Log mention, citation, position and — above all — the sources
For each prompt × engine, record: is the brand mentioned? cited with a clickable link? where in the answer does it appear? which competitors show up? And critically, which domains the engine cites as its sources. The source-domain column is the most valuable data you collect — it is the map of who the engines trust today.
- 4
Compute the metrics
Citation rate — answers that link the brand ÷ total answers — is the headline KPI. Also track mention rate (named but not linked) and share-of-voice (brand mentions ÷ all brand mentions in the set). Break each metric down per engine to see exactly where the brand is visible and where it is invisible.
- 5
Map the incumbents, surface the gaps
The most-cited source domains are who the engines trust right now — that list is your client’s digital-PR, outreach and comparison-content target list. The prompts where the brand never appears are the content backlog. The tracking data tells you precisely what to build and where to earn mentions next.
- 6
Report monthly and act on the trend
Re-run the identical prompt-set and engines every month, chart the trend, and convert the gaps into the next content sprint. A flat citation rate with a clear incumbent map is still a useful client report — it sets the baseline and lets you prove the work over time, which is what clients are actually paying to see.
One row per prompt × engine × month
| Column | What to record |
|---|---|
| run_date / engine | The month and which engine (chatgpt, perplexity, google-aim, gemini, copilot) |
| mentioned | Was the brand named at all? (Y/N) |
| cited | Named with a clickable source link? (Y/N) — the headline signal |
| position | Order it appears in the answer (1 = first brand named) |
| competitors | Which rival brands the answer named instead |
| source_domains | Every domain the engine cited as a source — your incumbent map |
| snippet | Verbatim text if the brand is described (sentiment + accuracy) |
Doing this across a client roster
The method is simple; doing it by hand across 10+ clients, every month, across five engines, is the bottleneck — and the part clients will pay for is the report, not the spreadsheet labour. The two halves that matter are producing the content that earns citations in the first place, and presenting the trend back to the client under your brand.
That's the pairing AltoRank is built around: an approval-first content engine that produces the answer-shaped, schema-rich pages AI engines cite — every page reviewed before it reaches a client's brand — plus a white-label reporting portal that surfaces GEO-citation tracking per client. The execution side for e-commerce is in optimize Shopify for ChatGPT Shopping; packaging and pricing it as an offer is in GEO services pricing.
Frequently asked questions
How do you track whether a brand is cited in ChatGPT?
By sampling. Freeze a set of buyer questions, run them through ChatGPT (and Perplexity, AI Overviews, Gemini, Copilot) in a logged-out session each month, and log whether the brand is mentioned, cited with a link, and which domains the engine used as sources. Citation rate — answers that link the brand ÷ total — is the headline metric you trend over time.
Is there a ranking API for ChatGPT or AI search?
No. Unlike Google rank tracking, AI answers have no positions API — you cannot query "where do I rank in ChatGPT." Visibility is measured by sampling a fixed prompt-set across the engines and recording mentions and citations. That is exactly why method consistency (same prompts, same engines, logged-out, monthly) matters so much: it is what turns a sample into a real trend line.
What is the headline metric for AI-search visibility?
Citation rate — the share of AI answers that cite the brand with a clickable link — is the headline KPI, because a citation is what sends trust and traffic. Support it with mention rate (named but unlinked) and share-of-voice (the brand’s mentions versus all brands in the set). Report all three per engine, since a brand can be strong in Perplexity and invisible in AI Overviews.
How often should agencies report AI citation tracking?
Monthly, using the identical prompt-set and engines every time. AI answers shift week to week, so a frozen monthly cadence is what separates a real trend from noise. Monthly also matches how agencies already report SEO, so AI-visibility reporting slots into the existing client cadence rather than adding a new one.
How do agencies do AI-citation tracking across many clients?
Per client, on a fixed monthly cadence, with the source-gap list feeding the content plan. Running it by hand across 10+ clients is the bottleneck — which is why AltoRank pairs an approval-first content engine (the answer-shaped, schema-rich content that earns citations) with a white-label reporting portal that surfaces GEO-citation tracking to each client under your brand.
The content that earns citations — plus the report that proves it
AltoRank produces answer-shaped, schema-rich content across every client store — reviewed before it ships — with a white-label portal to report GEO visibility back to each client.
Start the 14-day trialGive your agency its Friday afternoons back.
Start with one client workspace. Add the rest once you see the first article ship.