Free AI visibility check
See what ChatGPT, Gemini, Claude, DeepSeek and Perplexity actually say about your brand. Enter a URL, wait 1–2 minutes — we query the engines live, not a cache — and read the raw answers, scored out of 100 across 5 dimensions. No sign-up. 3 checks a day, free.
Two things worth knowing before you spend the 1–2 minutes: the report is written in Traditional Chinese and the probe prompts are generated in Cantonese, so the check is sharpest for brands selling into Hong Kong and Greater China. And how many engines answer depends on each provider's API that day — the report marks every engine as answered or unreachable, and unreachable ones are excluded from the score rather than guessed at.
Start a free check- Price
- Free
- Runtime
- 1–2 minutes
- Sign-up
- None
- Per day
- 3 checks
- Report language
- Traditional Chinese
Why check AI visibility at all?
You have tools for your Google ranking. What models say about you has been a blind spot. Three reasons it stops being one.
Buyers now ask an assistant before they ask Google
More and more shortlists start with "who's good at X" typed into ChatGPT or Perplexity. That happens before the search results page, where your ranking is invisible. If the model doesn't name you, you never reach the shortlist.
You can't edit what AI says about you — but you can read it
A model may describe you with three-year-old positioning, or confuse you with a same-named company elsewhere. You cannot patch a model's memory directly. You can see what it currently says: the check copies the raw wording back to you.
Without a baseline, optimisation is unprovable
Schema, BLUF rewrites and entity consistency take months to surface. With no starting score and no archived report, you can never show whether things improved or slid backwards. Take today's snapshot first.
What you get: 5 dimensions, 100 points
The score is not one opaque number. Each dimension has its own ceiling, so you can see exactly where the points leak.
Recommendation rate
35 ptsWhen the model is asked who does this kind of work well, does it bring you up unprompted? Weighted heaviest because it maps most directly to lost deals.
Description accuracy
25 ptsHow far the model's account of what you do and who you serve drifts from your actual positioning. Brands that renamed or repositioned bleed most here.
Brand awareness
15 ptsWhether the model knows you exist at all. At zero, nothing downstream can score — fix "never heard of you" before you argue about wording.
Sentiment
15 ptsWhether the mention reads positive, neutral or negative. Neutral dominates; negative usually means old news or complaints outrank your own content.
Citation sources
10 ptsWhether the model cites your site as a source. The hardest dimension to move, and the one that compounds most once it does.
Expect roughly ±10 points of natural variance — AI replies are inherently random. For a serious benchmark, run it three times and average.
And five things besides the score
AI engine matrix
6 prompt phrasings × each engine, as a red/amber/green grid. You can see at a glance which phrasing and which engine loses you.
Competitors the AI named instead
The rival brands the model volunteers when asked who's best — i.e. who is actually occupying your slot right now. (You don't enter their URLs; the model surfaces the names itself.)
Full raw reply to every prompt
Not just a score — the model's verbatim words about you, all six prompts, so you can audit where the number came from yourself.
Copy-ready content and schema
Generated FAQ entries, a brand description and Schema.org JSON-LD you can paste straight into your own site.
30-day action plan
Next month's sequence ordered by your weakest dimensions — not boilerplate advice.
Three steps, 1–2 minutes
- 1
Enter your URL
Paste your website address into the box. No account, no email, no card. 3 checks per day are free.
- 2
Wait for the engines to answer
Six questions about your brand go out to several major AI engines live, which takes 1–2 minutes. It's slower than a typical online checker because we are genuinely querying the models, not reading a cached index — keep the tab open.
- 3
Read the report
You get a 0–100 AI visibility score, the 5 dimension breakdown, each engine's raw reply, the competitors the AI recommended, copy-ready schema and a 30-day plan. The report has its own shareable link, and it is written in Traditional Chinese.
The tool does not flatter you
We ran it on ourselves: 41 → 43 out of 100
We sell AEO, so you are entitled to ask what we score. The answer is 2 points in 165 days, far short of the 70 we set ourselves. We also rebranded midway, which voluntarily wiped part of our accumulated AI memory and sent brand awareness backwards.
We published that number anyway, with all six verifiable reports and the dimension-by-dimension movement, caveats included. A tool that hands you 90-something is selling reassurance. A tool willing to publish 43 for its own maker is at least reporting a real number.
Read the full self-test case study (6 raw reports)Case study is in Traditional Chinese; the scores and report links are readable in any language.
Who builds this
Webka is an AEO (answer engine optimisation) agency based in Hong Kong. We build and monitor AI visibility for Hong Kong and Greater China brands — which is also why this tool probes in Cantonese and reports in Traditional Chinese. We built the check for our own client work first and left it free and open, because a baseline nobody can run is worth nothing.
Main site is in Traditional Chinese.
Frequently asked questions
Is it actually free, or is there a paywall on the report?
Actually free. 33 checks you wait for the next day. The report does carry two optional opt-ins at the bottom (a monthly trend report, or a 15-minute consult). If you leave an email there we will follow up about your report — that is how we find clients. Leave it blank and we never learn who you were.
How long does it take, and why 1–2 minutes rather than seconds?
1–2 minutes. We send six real prompts to each available AI engine and wait for them to finish answering, rather than looking your brand up in a pre-built database. That is the whole point, and it is also why it is slow — the exact time drifts with each provider's speed on the day. When it finishes, the report gets a permanent link so you never have to re-run it to read it again.
What language is the report in? Is there an English version?
The report is written in Traditional Chinese, and there is currently no English version — we would rather tell you that up front than have you wait 1–2 minutes to find out. The probe prompts are also generated in Cantonese, and the market frame defaults to Hong Kong, so the check is at its most accurate for brands that sell into Hong Kong or Greater China. You can still run any URL, and the parts that travel best regardless of language are the raw AI replies, the numeric scores, the competitor names and the generated Schema.org JSON-LD.
How accurate is it, and how many engines are tested?
The scores come from live answers the engines actually gave, not estimates or simulations, and every raw reply is laid out for you to check. The tool is wired to ChatGPT, Perplexity, DeepSeek, Gemini and Claude, but how many respond on a given day depends on each provider's API — the report lists each engine as answered or unreachable, and unreachable engines are excluded from the score rather than guessed at. AI replies are also inherently random: the same URL can differ by roughly ±10 points between runs, so for a serious benchmark run it three times and average (that is what we do on our own).
Can I compare myself against a specific competitor?
Not by entering their URL — there is no versus input. What you get instead is the set of rivals the AI itself names when asked who is best in your category, which is arguably the more useful list: it is who is taking your slot today. If you want a competitor's own full score, just run their URL as the subject of a check (it uses one of your 3 daily runs).
Go ask an AI: “who's the best [your industry] company?”
If you are not in the answer, you need to know where the gap is.
Free · No sign-up · 1–2 minutes · Report in Traditional Chinese
- 3 checks per day
- Shareable report link
- Names the competitors AI recommends