Where AI answers name you — across all seven engines

You have a rank tracker, so you know where you sit for your terms. Then a customer tells you they found you through ChatGPT, and you realise you have no idea whether that was luck. You open ChatGPT yourself, type something close to what they might have asked, and read the answer. Your name is not in it. You try a different phrasing and it is. Was that a win? You have no way to say.

So you do the honest thing and check the others. Perplexity gives a different answer. Gemini gives a third. By the time you get to Grok you have four browser tabs, no record of what you asked, and a growing suspicion that you are measuring your own mood rather than your visibility. Meanwhile the actual question is one you never thought to type, because real people do not phrase things the way keyword tools do.

What changed

AI Visibility now covers seven answer engines — Google AI Overview and AI Mode, ChatGPT, Perplexity, Gemini, Claude and Grok — asked the same terms in the same run. And a new Mentions tab searches an archive of real AI answers for your domain, so the questions come back as the result instead of something you had to guess first.

How it works in practice

Every engine answers on a credential you already have

The thing that usually stops people tracking seven engines is the bill. So each engine resolves its own route from what you have connected: DataForSEO, your own model API key with live web search, or a local CLI agent. A badge under each engine says which route it will use before you run anything — or names what to add when nothing can answer it yet.

Answers that come through your own key or a local agent cost nothing, so the pre-run estimate only charges for the engines that actually go through DataForSEO. The confirmation lists Billed engines and On your own keys side by side, so you agree to a number rather than discovering one.

One caveat the app states rather than hides: a local CLI answer has no web retrieval. It is the model's memory, labelled as such wherever it appears, and it is a baseline — not a live citation.

Find the questions you would never have guessed

The Mentions tab takes your domain instead of a term list and searches DataForSEO's archive of Google AI Overview and ChatGPT answers for it. Match on Cited as source — your domain is linked in the answer — or Named in answer, which catches prose mentions with or without a link. Subdomains are included.

Results roll up three ways: Questions, sorted by how often people ask them inside AI tools; Your pages, so you learn which URL actually got picked; and Brands, so you see who else is being named on the same ground. Open any question to read the archived answer, the ranked sources beside it, and which one is yours.

Then close the loop: send a question — or the top 20 in one click — into a new tracker and monitor it across all seven engines from then on. Every search is saved automatically, and reopening one costs nothing; it shows what the archive said that day rather than re-running.

Know whether a move is real

AI answers are not deterministic. Ask the same question twice and the brands named can change, which makes a single sample per term a coin flip wearing a metric's clothes.

A tracker can now ask each term up to five times per run, and the Engines tab reports each engine's visibility with a 95% range beside it. A change smaller than that range is noise. A 12% to 18% move inside a ±9pt band is not progress, and the table tells you so instead of letting you write it into a report.

A new Evidence tab holds every stored answer for the period with the engine, the route that fetched it, the model that wrote it, and which sample it was — so any number on the dashboard can be traced back to the text behind it. The shareable HTML report gained a By engine section carrying the same ranges, and a new alert fires when you are visible in one engine and effectively absent from another over the same terms.

Image Generation, live

Design Studio → Image Generation is no longer a coming-soon placeholder. Write a prompt, pick a size, and the result lands in your asset library.

Same routing idea as visibility: choose Local AI to run on a CLI subscription you already pay for — nothing metered per image — or API to use your own provider key. The prompt only leaves the machine on the API route.

Generate and History are separate tabs, and jobs are durable: they keep running across navigation and app restarts, so starting a render is not a commitment to stare at it. Completed outputs are ordinary Assets, with Download, Open in Assets, and Use as reference to seed the next one.

One small mercy: picking a CLI that has no image model behind it is refused up front, with the reason and a shortcut to one that works — instead of running a full generation and failing on a missing file at the end.

Before vs After

	Before	Now
Engine coverage	AI Overview and AI Mode	Seven engines, same terms, one run
Cost of extra engines	Every lookup billed	Your own key or local CLI answers for nothing
Finding the right prompts	Guess, then check	Search your domain; the questions come back
Reading a change	One sample, no error bar	Up to 5 samples with a 95% range per engine
Auditing a number	Trust the dashboard	Evidence tab with the raw answer, model and route
Campaign imagery	Prompt elsewhere, download, re-upload	Generate in Design Studio, saved straight to Assets

Who benefits most

If you sell to people who ask AI first. Developer tools, B2B SaaS, anything researched before it is bought. Your buyers are spread across ChatGPT, Perplexity and Google's AI Overview, and being cited in one tells you nothing about the other six.

If you report to someone. The 95% range and the Evidence archive are what turn "visibility improved" into a claim that survives a follow-up question.

If you have already been cited and do not know it. Mentions is the fastest way to find out — it needs nothing but your domain, and the questions it returns are the ones people actually ask.

Try it

Open AI Visibility, paste your domain into Mentions, and see which questions already name you. Then send the top 20 into a tracker and let all seven engines answer them on the credentials you are already paying for. You will find out what AI says about you — and, for the first time, whether the number moved or just wobbled.