To monitor what AI says about your brand, run the same buyer prompts through ChatGPT, Gemini, Perplexity and Claude on a schedule and log each answer.
Short answer: pick 15–25 prompts a buyer would actually ask, run each one in every assistant you care about, and record four things: whether you're named, who's named instead, what's said about you and which sources are cited. Repeat every week or two with the same prompts. The pattern over time matters more than any single answer, because AI responses vary from run to run.
Why does strong SEO not cover this?
Your SEO work still counts. Google says its AI Overviews and AI Mode rely on the same signals as normal search, with no extra requirements (Google Search Central). The problem is measurement. Your rank tracker shows positions, and your analytics show clicks. Neither shows the answer a buyer read in ChatGPT before deciding. Pew Research found that when Google showed an AI summary, users clicked a normal result in 8% of visits, compared with 15% without one (Pew Research Center). Fewer clicks means more of the decision happens where your dashboards can't see.
How do you set up AI visibility tracking?
Choose your prompts. Start with decision-stage questions: "best [category] for [use case]", "[competitor] alternatives", "[your brand] vs [competitor]", "how much does [category] cost". Write them the way a buyer would type them, not as keywords.
Choose your assistants. ChatGPT, Gemini, Perplexity and Claude cover most chat use. If Google traffic matters to you, add AI Overviews and AI Mode, and Microsoft Copilot if your buyers use Microsoft 365.
Keep conditions steady. Use the same wording each time, note the date, and use a fresh chat so earlier messages don't steer the answer. Note whether the assistant searched the web.
Log the answer. For each prompt and assistant, record: mentioned (yes/no), position, competitors named, how you're described, whether the facts are right, and cited URLs.
Mark the gaps. Any prompt where competitors appear and you don't is a gap. Log it in an AI recommendation gap template so you can rank which to fix first.
Should you track manually or with a tool?
Manual checks are fine for a first baseline. Past about 20 prompts across four assistants – 80 answers per round – a spreadsheet gets slow and inconsistent.
Option | Price (checked 5 Oct 2026) | What you get |
|---|---|---|
Manual (spreadsheet) | Free, plus your time | Full control; hard to repeat consistently |
Otterly.AI Lite | $29/mo, 15 prompts | Daily tracking of ChatGPT, AI Overviews, Perplexity and Copilot; Claude, Gemini and AI Mode are add-ons |
Semrush AI Visibility Toolkit | $99/mo, 25 prompts | Visibility score, daily prompt tracking, competitor prompt gaps |
blogawesome Starter | $49/mo | Tracking in ChatGPT, Gemini, Perplexity and Claude, plus research, drafting and publishing of 20 articles a month |
Sources: Otterly.AI pricing, Semrush, blogawesome. Tools that only track are cheaper if you already have a writing process; tools that also draft and publish save the hand-off.
How do you read the results?
Not named, competitors named: a content gap. Check what the cited sources have that you don't.
Named, but described wrongly: a facts problem. Fix old prices or features on your own site and on profiles the answers cite.
Named only in some assistants: look at the sources each one cites. Perplexity and Google lean on live results, so newer pages can show up there sooner.
Answers change between runs: normal. Judge by the share of runs that name you, not one result.
How do you close a gap once you find it?
Write the page that answers the gap prompt directly. Don't title it "Why [Your Brand] Is the Best…" – assistants favour sources that compare options fairly. A page like "Best project management tools for agencies under 20 people", with honest comparisons, real prices and a clear short answer, is far more likely to be cited. Then make sure it's indexable, link to it from related pages and re-check the prompt after two to four weeks.
If you're left out across the board, start with why it happens: why your brand disappears when AI answers a question. For a one-off baseline, see how to build a keyword-level AI recommendation audit in one afternoon.
Frequently asked questions
Which AI assistants should I track?
ChatGPT, Gemini, Perplexity and Claude for chat, plus Google AI Overviews and AI Mode if search traffic matters to you. Start with the ones your buyers use.
How often should I check?
Weekly or every two weeks for your most important prompts. Models and their sources change without notice, so a dated history is more useful than frequent one-off checks.
Can I fix a wrong or missing AI answer?
Often, over time, but nobody can guarantee it. Publishing a clear, accurate page that answers the prompt, and correcting your facts where answers draw them from, gives assistants something better to use.
To get a baseline before you set anything up, run the free AI visibility audit. It needs no signup, takes about 30 seconds and shows whether AI models recommend you, who they name instead and where your content gaps are.