Brand perception shows the qualities that AI answers tie to each tracked brand, with the phrasings found. Facts is a list of statements about your brand that a project checks AI answers against. A fact can be a price, the founding of the company or a feature.
Use brand perception to see how AI describes you next to your competitors. Use facts to see where AI answers contradict what you say about your brand.
What is brand perception?
Brand perception is the set of qualities that AI answers attach to a brand. Each quality is called an attribute. For every attribute, Depra AI also keeps the terms, which are the phrasings the answers used.
The numbers come from the latest Brand Perception read. The read states how many stored answers it covers. Your own brand and every tracked competitor are in it, so you can compare them.
How do I read the perception numbers?
Each row is one attribute for one brand. It has four fields.
| Field | What it means |
|---|---|
| Brand answers | The answers in the read that name that brand. This is the sample size of the row. |
| Hits | The answers naming that brand that also give it the quality. |
| Association | Hits as a percentage of brand answers. |
| Terms | The phrasings found for the attribute. |
Here is an example with made-up numbers. 40 answers name your brand, and 10 of them call it gentle on skin. The association for that quality is 25%.
Note: Read the brand answers count next to every percentage. A share over only a few answers is too small to read.
What are facts?
Facts are the statements a project checks AI answers against. You keep them in Project settings > Facts. Each fact has a kind and a text, and is either switched on or not.
A fact has one of 11 kinds. The list_facts tool names them PRICE, FOUNDED, HQ, FEATURE, INTEGRATION, AVAILABILITY, POLICY, CERTIFICATION, DATA_REGION, LANGUAGE and OTHER. HQ stands for headquarters.
The list_facts tool also returns the template slots that no fact fills yet, so you can see what is still missing.
What do supported and contradicted mean?
Supported and contradicted are two counts that Depra AI keeps for each fact, over the last 28 days. Both count claims, and a claim is something an AI answer states.
- Supported: the number of claims in AI answers that the fact supported.
- Contradicted: the number of claims in AI answers that the fact contradicted.
- Claims: the number of claims counted for that fact. It is the sample size, so read it first.
Here is an example with made-up numbers. Your PRICE fact says a product costs ₹499. For the last 28 days it shows 9 supported and 3 contradicted. So the fact supported 9 claims found in AI answers, and contradicted 3.
Depra AI shows the counts. It does not change an AI answer or your website. If a detail about your brand is out of date somewhere, the correction is yours to make: on your site, your marketplace listings and your social profiles.
What does "Build facts from your site" do?
"Build facts from your site" drafts facts from your own website. When you press it, DepraBot reads your homepage and up to 5 pages linked from it on the same site. These can be pricing, about, shipping, returns, contact or FAQ pages. DepraBot is the web reader behind Depra AI.
| Limit | Value |
|---|---|
| Pages read | Your homepage and up to 5 same-site pages linked from it |
| How often | At most once a day per project |
| robots.txt | Obeyed, as on every DepraBot request |
Read each drafted fact and correct it where needed. A page can be out of date, and a draft is only as right as the page it came from.
How many site fact drafts does my plan include?
Starter includes 2 site fact drafts a month, Growth 4, Scale 8 and Enterprise 16.
| Plan | Site fact drafts a month |
|---|---|
| Starter | 2 |
| Growth | 4 |
| Scale | 8 |
| Enterprise | 16 |
The daily limit applies on every plan: "Build facts from your site" reads a site at most once a day per project.
Can an AI assistant read perception and facts?
Yes, through the Depra AI MCP server, a read-only connection for AI apps. The get_brand_perception tool returns the attributes for each tracked brand. The list_facts tool returns each fact with its counts, and can be limited to one kind or to facts that are switched on.
The get_recent_changes tool also returns Brand Perception alerts from the last 7 days. These are entries in the data. Depra AI sends no alerts by email, WhatsApp or Slack.
See the MCP tools reference, or paste the facts prompt from MCP recipes.
What do perception and facts not do?
They show what AI answers say about your brand. They do not change it.
- No edits: Depra AI does not write content, change your website or change an AI answer.
- No guarantee: correcting a fact on your site does not guarantee that AI answers change.
- One window: supported and contradicted counts cover the last 28 days.
- Small samples: a share over only a few answers is too small to read.