An AI visibility audit checks how ChatGPT, Gemini, Perplexity, Claude, Microsoft Copilot and Google AI Overviews answer questions about your brand and your category, and whether they name you, skip you, or push a competitor instead. Here's how to run one properly in 2026, engine by engine.
- An AI visibility audit tests one prompt set across ChatGPT, Gemini, Perplexity, Claude, Copilot and Google AI Overviews.
- Koalr separates citations, mentions and absences, because AI engines treat them as different signals.
- Perplexity and Google AI Overviews pull fresh web results per query; ChatGPT's default mode leans on training data unless browsing is on.
- Build 30-50 real buyer questions before running any tool; a thin prompt set produces a false verdict.
- Best for marketing teams that need one view of brand visibility across all six major AI answer engines.
Why this matters
Ranking on Google page one used to be the finish line. Now a buyer asks ChatGPT or Gemini the same question and gets a narrated answer with two or three brand names attached — if yours isn't one of them, the click never happens. Koalr tracks that exact moment: whether an AI engine cites your brand, mentions it in passing, or leaves it out while naming a competitor.
An AI visibility audit is the diagnostic that tells you which of the six major engines see your brand and which don't. Skipping it means guessing at a channel that's already routing buyers away from your site in 2026.
How do you run an AI visibility audit across major AI platforms?
Run the same prompt set against every engine, log the outcome per prompt, and repeat on a schedule. The steps:
- Build a prompt set from real buyer language — 30 to 50 questions pulled from sales calls, support tickets and search queries, not guesses.
- Run each prompt against ChatGPT, Gemini, Perplexity, Claude, Microsoft Copilot and Google AI Overviews individually — a single-platform check misses where the actual gap sits.
- Log three outcomes per answer: cited (linked source), mentioned (named, no link), or absent (a competitor named instead, or nobody named at all).
- Score competitors on the same prompt set — visibility only means something relative to who else gets named in the same answer.
- Repeat monthly and track the trend, not the snapshot. One audit is a photo; recurring audits are a signal.
Citation behaviour by platform
| AI platform | Typical citation behaviour | What the audit should check |
|---|---|---|
| ChatGPT | Cites sources mainly when browsing/search mode is active | Whether turning browsing on changes the answer given |
| Gemini | Grounds answers in Google Search results | Whether you're a cited source, not just a passing mention |
| Perplexity | Cites sources on nearly every answer | Whether you're cited or excluded entirely |
| Claude | Cites less consistently; often names brands with no link | Whether it names you unprompted at all |
| Microsoft Copilot | Draws on Bing's index | Whether your Bing ranking matches your Google ranking |
| Google AI Overviews | Summarises top organic search results | Whether you rank organically but still miss the Overview |
The verdict: an AI visibility audit that only checks ChatGPT is checking one engine out of six that buyers actually use to shortlist brands in 2026.
ChatGPT: what to check
ChatGPT's citation behaviour depends on mode. Answers generated with browsing or search enabled show footnoted links; answers generated from the base model don't cite anything, they just name brands from training data. An audit has to run the same prompt both ways, because a brand can be cited in one mode and invisible in the other. Track your brand in ChatGPT and AI answers if this is the gap showing up most in your results.
Gemini: what to check
Gemini grounds many of its answers in live Google Search results, closer in behaviour to AI Overviews than to ChatGPT's base mode. The audit question isn't just "does Gemini name us" — it's whether Gemini treats your page as a cited source or drops your name into a list with no link attached.
Perplexity: what to check
Perplexity cites a source on nearly every claim it makes, which means a missing citation here is a real content gap, not a technical quirk. If your competitors show up cited and you don't, that's the clearest signal in the entire audit.
Claude: what to check
Claude cites less consistently than Perplexity or Gemini and will often name a brand with no source link at all. The useful metric here isn't citation rate — it's whether Claude names your brand unprompted when asked an open category question.
Microsoft Copilot: what to check
Copilot draws on Bing's index rather than Google's, so a page that ranks well on Google but poorly on Bing can be invisible to Copilot while doing fine everywhere else. Run the audit's prompts through Copilot separately — don't assume Google visibility carries over.
Google AI Overviews: what to check
AI Overviews summarise the top organic search results for a query, so a page missing from the Overview while ranking on page one usually signals a snippet-eligibility problem, not a ranking problem. That's a fixable, specific issue an audit should flag by name.
Why AI visibility audit results vary across platforms
- Data freshness — some engines browse live on every query, others lean on a training cutoff.
- Source indexing — Copilot pulls from Bing, most of the rest lean on Google's index.
- Citation behaviour — Perplexity cites almost every claim; ChatGPT and Claude cite less consistently.
- Prompt phrasing — a branded question and a category question return different answers on the same engine.
- Answer format — AI Overviews summarise search results; conversational engines synthesise across multiple sources at once.
- Model updates — engines ship new model versions without notice, and rankings shift between audits with zero change to your content.
“A citation drives traceable traffic; a mention doesn't — treating them as the same metric hides where the real opportunity sits.”
How often should you run an AI visibility audit?
Run an AI visibility audit monthly at minimum, weekly if you're actively publishing content to close a known gap. AI engines update model versions and underlying indexes without warning, so a quarterly check will miss real swings in whether your brand gets named.
What's the difference between a citation and a mention?
A citation includes a linked source pointing to your page; a mention names your brand with no link attached. The gap matters because a citation drives traceable traffic and a mention doesn't, so lumping them into one number hides where the actual opportunity sits.
Do you need separate prompts for each AI platform?
No — you need the same prompt set run separately on each platform, not a different set per engine. The same question phrased identically returns different brand names on ChatGPT versus Perplexity because each engine draws on a different index and citation style, and running one comparable set across all six is what makes month-to-month tracking mean anything.
Run your first AI visibility audit
See where ChatGPT, Gemini and Perplexity cite or skip your brand.
FAQ
What is an AI visibility audit?
An AI visibility audit tests a fixed set of buyer questions against major AI answer engines and logs whether each one cites, mentions, or omits your brand. It's the diagnostic step before any AI visibility or GEO work, run at least monthly through 2026.
Which AI platforms should an audit cover in 2026?
Cover ChatGPT, Gemini, Perplexity, Claude, Microsoft Copilot and Google AI Overviews. Each draws on a different index and cites sources differently, so skipping any one leaves a real blind spot.
How is an AI visibility audit different from an SEO audit?
An SEO audit checks rankings and technical health on Google; an AI visibility audit checks whether AI engines actually name your brand in a generated answer. A page can rank on page one and still never get cited by Perplexity or named by Gemini.
Can you run an AI visibility audit manually?
Yes, by asking each prompt in each platform's interface and logging the outcome in a spreadsheet, but it doesn't scale past a handful of prompts or a single month. A dedicated tool like Koalr automates the same prompt set across all six engines on a recurring schedule.
How many prompts should an AI visibility audit test?
Start with 30 to 50 prompts built from real buyer language — sales calls, support tickets, and search queries. Fewer than that and the audit misses category variation; far more and the analysis gets too noisy to act on.
Does ChatGPT cite sources the same way Perplexity does?
No. Perplexity cites a source on nearly every answer, while ChatGPT only cites when browsing or search mode is active and stays silent on sourcing in its base mode. An audit has to test both ChatGPT modes separately to catch the difference.
How often do AI answer engines change their citations?
Citations can shift between audits with no change to your own content, because engines update models and underlying indexes on their own schedule. That's why a one-off audit is a snapshot and a monthly one is the only version that shows a real trend.
Is Google AI Overviews part of an AI visibility audit?
Yes — Google AI Overviews summarises top organic results for a query, so it belongs in the audit alongside conversational engines. A brand can rank organically and still get skipped by the Overview, which is a distinct, fixable gap worth tracking on its own.
One last thing
The biggest visibility gaps rarely trace back to your own website. They show up when an engine cites a competitor's comparison page or a third-party review site instead of you — which means the fix in 2026 is often earning coverage elsewhere, not rewriting a page you already control.
