For the complete documentation index, see llms.txt. Every page on this site is also served as Markdown: append `.md` to any URL, or send `Accept: text/markdown`.
Skip to content
Attensira LogoAttensira
Book a demo

Which AI engines Attensira reads, and how

Attensira reads the AI engines your buyers use: ChatGPT, Perplexity, Claude and Google AI on Starter, plus Gemini on Growth and up. Each engine keeps one stable name when its vendor swaps the model, so history never splits. An engine that is off or cannot be read shows as not measured, never 0%. Copilot and DeepSeek are coming.

Start freeBook a demo

7 days free · card required

Settings · Enginesattensira.com
Engines read in this workspace
ChatGPT2,346 answers read in the window
Tracked
Perplexity2,172 answers read in the window
Tracked
ClaudeSwitched off here: never asked, so never 0%
Not tracked
Google AIConsumer reading only: there is no API reader
Not tracked
Copilot, DeepSeekNot read yet, so no number is shown
Coming

Two engines are switched on in our own workspace, so Claude and Google AI show as not tracked, not as 0%. Switching one on is a setting, and its history starts that day.

Shown as a baseline, not a result.

Analyst agentAnalystattensira.com workspace, read over the Attensira MCP · · 30 days to 2026-10-04, 24 of 31 days sampled

  • Analyst agentAnalyst · Reads each engine on its own terms, and labels any engine it could not read instead of scoring it.

How are the engines read?

An engine is the surface a buyer uses, not a model version. Attensira records how each was read and never presents one kind of reading as another.

  1. Start on the default set

    A new workspace reads ChatGPT, Perplexity, Claude and Google AI. Gemini is switched on by hand on Growth and up. Switching it on in Starter changes the setting, not the readings.

  2. Read once a day, per engine

    Each prompt, engine and country is read at most once a day, one answer per reading on Starter and three on Growth and up. A failed read lowers n, never your rate.

  3. Consumer or API, said plainly

    ChatGPT, Perplexity and Gemini can be read as the consumer product and over the API. Claude is read over its API. Google AI has no API reader, so its data comes from consumer reading only.

  4. Choose wide or deep

    Every engine you add multiplies the readings your plan's responses must cover. Read four or five engines over fewer prompts to find where buyers are, or two engines over many prompts once you know.

Do two engines treat the same brand the same way?

No. ChatGPT links our pages more often than it names us; Perplexity names us a little more often than it links us. Each engine keeps its own rates, side by side, never averaged.

named · ChatGPT
4.2%
n=2,346 answers
linked · ChatGPT
6.6%
n=2,346 answers
named · Perplexity
4.5%
n=2,172 answers
linked · Perplexity
3.9%
n=2,172 answers
Analyst agentAnalyst · Kept each engine apartattensira.com workspace, read over the Attensira MCP on 4 Oct 2026 · 24 of 31 days sampled · to

What do tracked, readable and zero mean?

Every per-engine number carries flags. Reading them wrong is the most common way to misreport a result.

What do tracked, readable and zero mean?
Not trackedThe engine is not switched on in this workspace. No question was asked, so there is no answer and no zero.
Not readableNo reader exists for that surface, such as Google AI over the API. No setting turns it on.
Not measuredTracked, but no successful answer came back in the window.
0% with n = 12Asked twelve times and never named you. A real, measured zero.
A model is not a crawlerAn engine answers a prompt; a crawler fetches your pages. A high crawl rate does not produce mentions. Crawlers are reported separately, in crawler logs.
Stable nameschatgpt, perplexity, claude, gemini and google_ai keep their id when a vendor swaps the model underneath, so your history never splits.

What does Attensira not read?

The engines read are defined in the product, not by the list of assistants that exist. Anything else is named only as not read.

  • Copilot and DeepSeek are not read yet, so no number is ever shown for them.
  • A newly added engine does not backfill. Its history starts on the day you switch it on.
  • Turning an engine off leaves a gap in its series, shown as not tracked, not as zero.
  • Google AI data is thinner and less regular than the API-backed engines. Say so in any report built on it.

What does one engine say about tracking another?

Perplexity on 2 Oct, asked for a “ChatGPT rank tracker”: mentions, citations and visibility, not a search ranking. Each engine is read on its own, so this answer counts for Perplexity and never for ChatGPT.

Answer · Perplexityasked

Question: ChatGPT rank tracker

Perplexity answered: ChatGPT rank tracking tools measure how often and where your brand appears in AI-generated responses, tracking mentions, citations, and visibility rather than traditional search rankings.

Sources cited

  1. 1rankability.com
  2. 2rankscale.ai
  3. 3aiclicks.io
  4. 4geoptie.com
  5. 5morningscore.io
  6. 6useomnia.com
  7. +4 more

Not namedAttensira is missing from this Perplexity answer.

Researcher agentResearcher · Stored the answerattensira.com stored Perplexity answer, read over the Attensira MCP · · 10 links

What else do people ask about AI engines?

Which engines can I read on each plan?
Starter reads ChatGPT, Perplexity, Claude and Google AI every day. Growth and Business add Gemini. Answers can be read country by country in the markets you add. Copilot and DeepSeek are coming soon.
Does Attensira track Google AI Overviews?
Yes, as Google AI, on every plan. It is read as the consumer product, because Google offers no API reader for it, so expect thinner data than on the API-backed engines.
Why not read every engine on every plan?
Every engine multiplies the readings: prompts × engines × markets × samples a day. The plan's monthly responses set how much can be read, and you choose how to spend them.
What happens when OpenAI ships a new model?
Nothing breaks. The engine keeps its stable name, chatgpt, so your history continues in one series instead of splitting at every release.
How long before a new engine's numbers mean something?
About two weeks. Rates on a new engine are noisy until n builds, and the 95% significance test will refuse to call a change until it has the evidence.

Read the engines your buyers ask, starting today.

Start freeBook a demo

7 days free · card required