> ## Documentation Index
> Fetch the complete documentation index at: https://help.aibuku.com/llms.txt
> Use this file to discover all available pages before exploring further.

# How AI talks about you

> See how each AI assistant speaks of your business, what it says on each topic, and where a bad statement comes from.

<Note>
  This is part of **Search presence**, under [Visibility](/visibility/overview), which some plans include.
</Note>

AI assistants describe your business to buyers in their own words. **How AI talks about you** shows how each assistant speaks of you, what it says on each topic, and where a bad statement comes from.

Open **Search presence**, then the **How AI talks about you** tab.

## How AI speaks of you

Each assistant has its own row. We never add assistants together: each one answers differently.

A row shows a score from −100 to +100 over the last 4 weeks, for all your questions. The score is the answers that speak well of you, minus the answers that speak badly, out of all the answers that name you. Next to it you see the answers it comes from, for example *21 good, 12 neutral, 5 bad of 38 answers*, and where they came from: answers we asked live, your own test runs, and an index of AI answers (Google's AI Overviews among them).

* An answer that does not name you is not in any score. Whether AI names you is on the overview.
* Fewer than 10 answers: you see **Not enough answers yet** with the number so far, and no score.
* The trend compares the score with the 4 weeks before. It says *Up* or *Down* only when the change is bigger than chance; otherwise *Steady*.
* The small line shows the score week by week. When an assistant changes, the line restarts and says so, for example *Gemini changed on 20 Aug*. Scores from before and after a change are not compared, so the trend says *No trend yet*.

## What each assistant says, by topic

The grid counts the statements of each assistant on each topic that speak well or badly of you. One answer can make several statements. Press a number to read them, word for word.

Next to a bad statement you see where it came from:

* *Comes from* a site: a page the assistant cites says the same.
* *Not what* a site *says*: the page it cites says otherwise.
* *The assistant's own view*: nothing it cites covers it.
* *Not checked*: we could not check it yet.

Only topics our checker reads reliably are shown. More topics appear once it reads them reliably too.

## What to do

* A bad statement that comes from another site: ask that site to correct it, or answer it on your own pages.
* A bad statement that is the assistant's own view: make sure your website states the facts clearly. Assistants change slowly; we check again every week.
* The fixes these answers raise are also listed in Checks & fixes.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.