Run a fixed set of real buyer prompts, signed out, across ChatGPT, Perplexity and Google AI Overviews, and record whether your brand is cited, named or missed. Run the same prompts more than once because answers vary, and log blocked checks honestly rather than guessing. A readiness scorecard checks your site's foundations, yet it never replaces live prompt tests.
Typing your own company name into ChatGPT tells you very little. A model will happily summarise what it knows about you from LinkedIn or an old press release while leaving you completely out of real buying conversations.
Build a fixed prompt set from discovery calls, support questions and competitor comparisons. For a B2B team, that means category and shortlist wording, not brand vanity queries. Keep the wording locked from month to month so you are comparing apples with apples.
Signed-in chats lie to you. If you have spent all month visiting your own website and asking ChatGPT about your product, personal memory and browser context can nudge your brand into the answer. Open a fresh incognito window, stay signed out where the engine allows it, and test ChatGPT, Perplexity and Google AI Overviews separately.
They will not give you the same result. On 26 Oct 2026, logged-out ChatGPT cited simonhawtin.com on 4 of my 5 weekly prompts, while Google's AI Overviews and Copilot were both 0 of 5 on those same prompts. Checking only one engine misses half the picture.
Keep the scoring simple so everyone reads it the same way:
I also log technical walls rather than pretending they are passes or fails. On 2 Nov 2026, Google hit an unusual-traffic block on query one, so that week's Google column says blocked. On 9 Nov 2026, Google returned 0 of 2 measured before blocking on query three, while ChatGPT held at 3 of 5 (down from 4 of 5 on 26 Oct) and Copilot stayed 0 of 5. Keeping that dated log stops you fooling yourself when numbers wobble. You can see the full weekly record in the agentic AEO case study.
Manually running 30 or 50 prompts every week gets boring fast, and automated browser checks get blocked by captchas, as my Google log shows. Once you know which buyer questions matter, a dedicated visibility tracker such as Peec gives you a steadier daily or weekly baseline across engines without relying on ad-hoc scraping. I would still keep signed-out manual checks for spot checks when something changes, not as your only system of record.
I have not run a bake-off across every visibility tool on the market, so I am not going to publish a ranked "best tools" league table I cannot back up. Pick a tracker that covers the engines your buyers use, tracks both brand mentions and source URLs, and lets you keep your prompt wording fixed.
Prompt visibility is only the top of the funnel. Pair your prompt log with AI referral tracking in analytics and a "How did you hear about us" field on enquiry forms that lists AI assistants. A buyer who shortlists you in Claude or ChatGPT often arrives later via branded search or direct, so self-reported attribution catches what referral headers miss. For how to turn those visits into enquiries, see how to turn AI search traffic into leads, and check your foundations on the Answer Engine Readiness Scorecard.
Weekly or daily if you use a tracker, and always using the exact same prompt wording. Answers vary between runs, so one check on one afternoon is not a reliable trend.
At a minimum, check ChatGPT, Perplexity and Google AI Overviews, because buyers use them differently and they rarely cite the same pages. Checking signed out avoids personal history skewing the answer.
The scorecard checks whether your pages, entity clarity and technical basics are ready to be cited. Checking visibility means running live buyer prompts and scoring whether engines actually cite, name or miss you.
Log it as blocked rather than guessing or repeating until you hit a captcha. On 2 Nov 2026 Google blocked my check on the first query, and on 9 Nov it blocked from the third query onwards, so I recorded the gap plainly.
No. You can run a small fixed set of 5 to 10 buyer prompts signed out in a spreadsheet first. Once you want steady trend lines across 30 or more prompts without browser blocks, a dedicated tracker makes life much easier.
Free discovery call
A free 30-minute call. Together, on the call, we look at whether buyers in your category use AI to shortlist, whether you show up, and whether you can act on the fixes. If it isn't worth doing, I'll tell you. The £595 audit and the 3-month service both start here.