An AI visibility score is built from what AI assistants actually said about you. It is not a readiness check, and the difference decides what you should work on next.
Last updated: August 18, 2026 - Reviewed by the DataEase AI editorial team
The phrase AI visibility score gets attached to two different measurements, including on this site, and confusing them costs teams months. One is a readiness check you can run this afternoon, before a single AI assistant has heard your name. The other cannot exist until real answers have been collected.
This page is about the second one: the AI Visibility Score inside DataEase AI, measured from what ChatGPT, Claude, Gemini, Perplexity, and Grok actually said about your category. It covers the 4 pillars behind it, what a good number looks like, and where the free readiness score fits instead.
An AI visibility score is a 0 to 100 measure of what AI assistants actually said about your brand when asked about your category. DataEase AI builds it from 4 pillars: mentions, recommendation, citations, and sentiment, in that order of weight.
The load-bearing word is said. This is not a forecast, a crawl of your own website, or a proxy built from backlinks. It is a record of collected answers: questions get asked, answers come back, each answer is parsed for named brands, and the score summarises what was found. Before any answers exist, there is no score, and a tool that hands you one has computed something else.
Read as a record rather than a grade, a score of 12 is not a judgement about your product. It is a count of category answers that went past without your name in them, which is a much more actionable fact.
The wider concept, including the public signals that make a brand citable at all, is covered in our pillar on AI brand visibility and how to raise it. This page is about the number itself.
Brand readiness scores whether you are set up to be found, across 7 pillars, before anyone has mentioned you. AI visibility scores what up to 5 AI assistants actually said. Readiness is a cause you control today; visibility is the recorded effect.
| Brand Readiness | AI Visibility Score | |
|---|---|---|
| Question it answers | Are you set up to be found and trusted? | Were you actually named in the answer? |
| Evidence it reads | Your own public surfaces: name, site, profiles, markup, trust pages | Answers collected from AI assistants when asked category questions |
| Structure | 7 pillars, from Name Quality to Brand Authority | 4 pillars: mentions, recommendation, citations, sentiment |
| Can you run it cold? | Yes, on day 1, before any mention exists anywhere | No. It needs collected answers to exist at all |
| How it is read | Bands from Early Stage 0-55 up to Iconic 91-100 | Read against the 5 Brand Presence tiers, Invisible up to Authoritative |
| Where you get it | The free Brand Analyzer, no signup | The DataEase AI Branding app, after a scan |
The relationship is causal but loose. Readiness makes you eligible: a brand whose name collides with three other companies, whose site carries no structured data, and which nobody outside its own domain has written about will not be named. The full pillar-by-pillar definition is in our glossary entry on brand readiness and its 7 pillars.
But readiness does not buy mentions. Scoring well on all 7 readiness pillars while staying invisible in category answers is normal, because incumbents already occupy them and the models were trained before you existed. That gap is the cold-start problem. One number tells you the door is unlocked; the other tells you whether anyone walked through it, so run readiness first, fix what it flags, then start collecting answers.
The score has 4 pillars, ranked by weight. Mentions counts how often you are named. Recommendation counts how often you are actively recommended. Citations counts distinct high-authority sources about you. Sentiment counts how positively you are described, and carries the least weight.
| Pillar | What it measures | Relative importance |
|---|---|---|
| 1. Mentions | How often your brand is named at all across the collected answers for your category | Carries the most weight. Everything else scales off it |
| 2. Recommendation | How often you are actively recommended rather than merely listed among options | Second heaviest. The difference between being an option and being the answer |
| 3. Citations | How many distinct high-authority sources are cited when assistants talk about you | Third. Corroboration is what makes a mention repeatable |
| 4. Sentiment | How positively you are described in the answers where you appear | Lightest, and capped. Counts only in proportion to how often you are mentioned |
Note what is absent: no pillar for followers, published word count, or how good your site looks. Every pillar is a property of an answer that was produced.
We do not publish the weights, and that is deliberate. Published weights get gamed rather than earned, and the score stops describing reality. What we will always tell you is the ordering, because that is what you need to sequence work: mentions first, recommendation next, citations after that, sentiment last and capped.
Inside the app these four are not just a rollup. The AI Visibility section breaks out Mentions, Sentiment, Citations, Partnerships, Search overlap, History, Market share, and Head-to-head, so a movement in the composite traces back to the pillar and the question that moved it.
Because sentiment only counts in proportion to how often you are named. A brand appearing in 2 of 40 collected answers earns almost nothing from sentiment even when both mentions are glowing, and sentiment is capped, so it can never carry a score on its own.
This is the most useful thing to understand about AI visibility measurement, and it inverts the instinct most teams bring from social listening. There, a wave of positive posts is the win condition. In AI answers, sentiment is nearly worthless on its own, because the answer that never named you contributed a silent zero.
Two brands, same category questions. The first is named in most answers, described neutrally, recommended sometimes. The second is named twice, both times in warm, flattering language. The first is winning by an enormous margin, and any scoring model that lets the second look competitive is lying to its user. Positive sentiment is worth nothing to a brand AI assistants rarely name.
The cap enforces that. Without it, a brand could grind sentiment upward and watch a composite climb while its real presence stayed flat. So if your mentions pillar is low, do not spend a week on tone or testimonials. Spend it on being present at all, which is content and entity work, covered in our guide to answer engine optimization.
A citation is a distinct source an AI assistant referenced when it talked about you. DataEase AI counts distinct high-authority domains rather than raw link counts, so 6 references to your own homepage inside one answer still resolve to a single source.
Distinctness is the whole point. A model quoting your own marketing site five times has one reason to believe you exist, not five. A model quoting a trade publication, a comparison roundup, and a substantive community thread has three independent reasons, and independence is what survives a re-crawl. Counting raw links would reward volume on a domain you already control.
Authority matters as much. A directory nobody reads and an established industry publication are not equivalent evidence, and treating them as equivalent would make the pillar trivially inflatable. Citations explain why you were named; being named is still the thing being measured.
The crawl side of the same story is tracked separately. AI crawler monitoring covers 28 AI crawlers across today, 7, 30, and 90-day windows, with recrawl cadence per page and coverage gaps. One busy brand logged roughly 335,000 crawler visits in 30 days. A page no crawler has ever fetched can never become a citation.
By asking real questions and recording real answers. DataEase AI runs your category questions across up to 5 AI assistants, with ChatGPT, Perplexity, and Gemini on every plan and Claude and Grok on Business, then parses each answer for the brands it named.
The question set comes from discovery, which maps your competitors, your personas, and the questions those buyers actually ask. Discovery is free and consumes no credits, which matters because the question set is the measurement instrument. Questions are grouped into 6 types - Discovery, Comparison, Recommendation, Pricing, Reviews, and Use Cases - so winning recommendation questions and losing pricing questions is visible rather than averaged away.
Branded and direct questions are excluded. An assistant naming you when the question already contains your name proves nothing, and including those questions is the easiest way for a tool to produce a flattering chart. What is measured is genuine category demand.
Parsing is harder than it sounds. AI assistants routinely drop part of a name, and an answer recommending a brand while omitting the suffix in its registered name scored zero mentions until brand aliases were supported. With an alias registered, that same sentence counts as one mention and is flagged as a recommendation. Discovery now proposes aliases for you and for every competitor.
The same answers feed market share of AI answers, a rank-weighted share of your category. Being named first counts for more than being named tenth, and the credit for each answer always adds up to one. It is reported by brand, by assistant, across the last 6 scans, and by question type.
There is no public benchmark for the AI Visibility pillar in isolation, so read it against the rollup. DataEase AI sorts Brand Presence Scores into 5 tiers: Invisible 0-20, Emerging 21-40, Established 41-60, Strong 61-80, and Authoritative 81-100.
| Brand Presence Score | Tier | What it means in practice |
|---|---|---|
| 0-20 | Invisible | Category answers happen without you. Normal for pre-launch and newly launched brands |
| 21-40 | Emerging | You surface occasionally, usually in long-tail or niche questions rather than headline ones |
| 41-60 | Established | Named reliably in your category, but usually listed as an option rather than recommended |
| 61-80 | Strong | Named often and recommended regularly, with independent sources backing the mentions |
| 81-100 | Authoritative | The default answer in your category, cited from sources you do not control |
Two clarifications, because these bands get misquoted. These are the Brand Presence tiers, the rollup of 5 components, and AI Visibility is one input to it rather than the whole. They are also not the readiness bands, which run Early Stage 0-55 through Iconic 91-100 on a scale that does not map onto this one.
Most early-stage companies land in Invisible or Emerging. That is arithmetic rather than failure: the models were trained on a world that did not contain you, and the incumbents have a decade of corroborating sources.
The number that matters is the trend against your own named competitors. Moving from 18 to 31 over a quarter while the leader sits flat is the right work, even when the absolute score still looks bad. That is what the head-to-head battleground compares: four slots, your own brand locked into the first, question by question.
Nothing is judged before 7 days. DataEase AI measures outcomes as matched pairs, comparing the N days after a fix against the N days before it, across visits, AI crawls, citations, indexing, and page score. Meaningful movement usually takes 4 to 8 weeks.
The matched-pair design exists because the naive alternative misleads. Looking at the week after a fix in isolation catches every seasonal bump, every unrelated launch, and every random fluctuation in phrasing. Comparing equal windows on either side of the change is much harder to fool yourself with, and the 7-day floor stops anyone celebrating a Tuesday.
There is deliberately no predicted score lift anywhere in the product. Two estimators were built and then deleted, because the numbers were invented, and an invented number becomes the thing a team plans around. You see impact and effort on each opportunity, then the measured outcome.
The loop is cheap by design. Findings live in a filterable table, and marking one as fixed moves it to a Tracking tab. Re-auditing a single page costs 1 credit instead of re-running the full 40-page crawl and returns in about 6 to 12 seconds, and self-fix detection verifies up to 12 pages per run. It suggests, it never auto-applies. Judge the first month on whether the inputs moved, not on whether the score did.
Inside the Branding app. Discovery finds your competitors and the questions buyers ask without consuming credits, and every scan shows its exact price before it runs. You get 100 free credits the moment you add a brand domain, with no credit card and no trial timer.
The Branding app is the core surface for Brand Presence Intelligence, and this is one of the three scores it maintains alongside Brand Readiness and the Brand Presence rollup. Supporting capabilities feed the same loop: Pages ships the content that closes a citation gap, and FormsAI responses appear automatically in your Dashboard, so a lead that arrived from an AI answer lands next to the visibility data that produced it.
Collection runs on the agent workforce. The Visibility Monitor watches ChatGPT, Perplexity, and Gemini, Reputation Watch scans mentions and sentiment, and the Content and Directory agents close the gaps that Opportunity Scout finds. Low-risk work ships on its own and high-impact changes wait for your approval, so it works autonomously while you stay in control.
After each scan, a review answers five questions with clickable evidence: why you were cited, why competitors were preferred, what changes your citations, what your last changes did, and what to do next. The model writing it is forbidden from producing a number of its own, so every figure is materialised by the server.
On cost, the honest version: no perpetual free tier, but no card and no timer either. Pay As You Go is 50 dollars per 100 credits, Growth is 249 dollars per month with 600 credits and 3 AI assistants, and Business is 449 dollars per month with 1,500 credits and all 5. Pack credits roll over, paid schedules pause and auto-resume rather than failing, and a failed run refunds its credits. Full detail is on the DataEase AI pricing page.
Add a brand domain and get 100 free credits. Discovery is free, every scan shows its price before it runs, and you keep the evidence behind every number.
Start with 100 free credits ->100 free credits. No credit card, no trial timer.