Six AI engines: why each one answers differently

Ask the same question on six AI engines and you get six answers. What each engine is, why they differ, and how to read your standing engine by engine.

By the BotSay team26 Oct 20266 min read

On this page

The six engines, in plain terms
Why the same question gets different answers
See where you stand on each engine
When one engine names you less
Two engines that run a day behind
A first check by engine, by hand
Let the engines tell you where to work

Put the same buyer question to six AI engines and you will read six different answers. One names you first, another names you third, and another leaves you out and lists two rivals instead. Each engine is a separate place where a buyer can meet your brand, and each can be read on its own. This post explains what the six engines are, why their answers differ, and how to use the differences to decide where to work first.

The six engines, in plain terms

BotSay asks every question on six engines: ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews and Google AI Mode. They are not six copies of one product.

ChatGPT is OpenAI's assistant. It answers from its own knowledge and can search the web.
Perplexity is an AI search engine. It reads pages and answers with citations to them.
Gemini is Google's assistant, asked directly. It is separate from what Google search shows.
Claude is Anthropic's assistant. It answers from its own knowledge and can search the web.
Google AI Overviews is the AI summary at the top of an ordinary Google search results page.
Google AI Mode is Google's separate AI chat tab on search, and it is not the same as Google AI Overviews.

Why the same question gets different answers

Several things shape what an engine says, and they vary from one engine to the next.

Where the answer comes from. An assistant can answer from what it already knows, or it can search the web first and write from the pages it finds. Perplexity reads pages and cites them as a matter of course. Google AI Overviews appears at the top of a search results page. That is one reason the brands named can differ from engine to engine.

What the engine searched for. When an engine searches before it answers, the words it searches with decide which pages it reads. Two engines can turn the same question into different searches and land on different pages. A review site that lists you may come up for one engine's searches and not another's.

Whether there is an answer at all. Google does not show an AI Overview for every search. Some of your questions will have no overview to read in a given week, which is normal for it.

The day you ask. Answers shift from day to day even when nothing about your brand has changed. One reading on one engine is a snapshot, so ask the same questions in the same words every week and compare like with like.

See where you stand on each engine

The simplest way to start is to split your numbers by engine. Mention rate, the share of answers that named you, is the place to begin. Work it out for each engine, then for all six together.

Mention rate, by engineAll engines: 34.0%

ChatGPT41.3%
Perplexity37.8%
Google AI Mode33.4%
Gemini33.2%
Claude29.9%
Google AI Overviews28.4%

Example, using an invented brand

Expect the engines to differ, and use it. The strongest engine shows what is already working, and the weakest shows where a rival is getting the mention you could have. Write the number of answers beside each engine's figure, because an engine that answered fewer questions gives a figure that can swing further on a single reply.

In BotSay, the By engine tab on Standing does this for you. It shows your Mention rate, Share of voice, Average rank, Your pages cited and Sentiment on each engine that answered. Beside them is how many answers that engine gave. The highest engine is marked strongest and the lowest weakest. The last row, All six, blended, is your figure across every engine, and it is the same figure the This week page shows, so the two screens always agree (Compare yourself with rivals on Standing).

When one engine names you less

When one engine names you clearly less than the others, BotSay adds a card for it on the By engine tab, named for that engine, such as Why Perplexity is low. The card gives a count: on how many questions that engine named a rival and not you. Its Read button opens Evidence filtered to that engine, so you go straight to what it said.

The card is a count. It tells you where to look, and the answers tell you why.

Read what the engine says. Which rivals does it name, and what does it say about them?
Open the pages it cited. If the same review site, publisher or community thread keeps coming up, that is a place to be listed or to take part.
Look at what it searched for. For every answer, BotSay shows the searches the engine ran to write it, or a note when an engine does not report them (What the Evidence page shows).

Perplexity, the searches behind one answer

What is the best payroll software for a 20-person company?

best payroll software for small businesses
payroll software with multi-state tax filing
Tessera vs Novaly pricing

Its answer cited 7 pages.

Example, using an invented brand

The searches tell you the words the engine went looking for. If none of your pages answers those words plainly, that is the page to write. If a rival's page does, read how it is built and write a better answer to the same question on your own site.

Two engines that run a day behind

Google AI Overviews and Google AI Mode reach BotSay through licensed search data, which comes back about a day after the other four engines. ChatGPT, Perplexity, Gemini and Claude answer in the same update that asks them. Until the two Google engines arrive, figures that can still move carry a Provisional mark (Which AI engines BotSay asks).

Because Google AI Overviews does not show a summary for every search, a reading can hold fewer answers from it than from the others. BotSay works out Mention rate over the questions an engine did answer, and the Readings page in Setup says so.

A first check by engine, by hand

To see the differences before you track anything, ask ten buyer questions on all six engines and log each answer in its own row, with the engine beside it. For Google AI Overviews, search the question on Google and read the overview when one appears. Then count by engine before you count across all six. Sixty answers show which engines know you and which favour a rival, though they are too few to tell a small move from ordinary variation. How many questions do you need to track AI visibility? shows how the number of questions changes what you can say, and our step-by-step guide to checking what AI says about your brand covers the rest of the hand check.

Let the engines tell you where to work

Treat the six engines as six readings of the same market. Where all six name a rival and not you, the gap is about the topic, and a page that answers the question well is the likely fix. Where only one engine leaves you out, start with the pages and searches behind that engine's answers.

BotSay asks your questions on all six engines every week, through each provider's official developer access or a licensed search-data provider, with web search switched on where the engine offers it. Every reading asks every question on every engine, so any two weeks compare like with like. All six engines are on for every plan. Your reading arrives in the weekly email, and Standing has the engine-by-engine detail when you want it.

What you gain with BotSay

All six engines on every plan, asked the same questions every week.
Your five measures for each engine, with the strongest and weakest marked and a blended figure across all six.
A card when one engine names you clearly less than the rest, with a count and a button that opens its answers.
The searches behind each answer, where the engine reports them, so you can see the words each engine went looking for.
One fixed monthly price with everything included and no add-ons. Plans differ only in the number of questions, from $59 a month for 60 questions.