DiscoveryHQ
Loading…
DiscoveryHQ
Loading…
Features
Give us your website address. We ask ChatGPT and Gemini the questions your customers ask, read your site the way AI reads it, write down who gets recommended instead of you — and then write the fixes. Every number opens onto the answers behind it.
2
AI assistants tested
ChatGPT and Gemini, through the AI companies’ own official services with web search turned on. Those two, and only those two.
3
Tries per question
AI answers move around. Asking once is a coin flip, so every tracked question is asked three times on each assistant.
18
Checks in the free check-up
3 questions × 2 assistants × 3 tries. It takes about one to two minutes. Free to start, no card.
Nothing runs on a schedule. There is no monitoring, no daily tracking and no alerts. A check-up happens when you choose to run one, and never in between.
01 · Measure
One number from 0 to 100: how often AI recommends your business, and how high up your name comes.
We ask what your customers actually type — “who’s the best plumber near me?”, “what should a small team use for this?” Not “do you know my business?”, which AI answers politely and tells you nothing.
The same question can get different names on different days. Three tries on ChatGPT and three on Gemini turns one lucky answer into something you can count on.
You see the tally, in the same words every time. There is no confidence figure on your report, because no honest one exists.
Recommended in 2 of 9 checks
That is what a result looks like — a count you can open and read.
Most people read the first name and stop. So the score follows the order AI put you in: named first is worth more than named third, and both beat not being named.
02 · Website analysis
Before AI can recommend you it has to reach your pages, understand what you sell, and trust what it finds. We check all three.
We walk your site the way an assistant does — following your links, reading what is actually on the page. Inside the app, a scan covers up to 100 pages.
0 to 100, from a fixed checklist: same site, same answer, every time. The list adapts to your kind of business — a restaurant is checked on opening hours and reviews, a software company on pricing and comparison pages.
We collect the questions customers ask in your line of work, then count how many of them your site answers. “You answer 8 of 42” is a to-do list, not an opinion — and the 34 are listed for you.
There is a small code label that tells AI what your business is — your name, what you sell, where you are, when you are open. Visitors never see it; AI reads it first.
If it is missing or wrong, AI has to guess from your wording, and a guess often lands on somebody else. We check whether yours is there and whether it says the right thing.
03 · Competitors
Every time an assistant answers without you, somebody else got the recommendation. We write down who, how often, and why.
Every business the assistants recommend for your questions, gathered across all the checks — including the ones you had never thought of as competitors.
Counted exactly the way you are counted: same questions, same assistants, same three tries. So the comparison is like for like, not a feeling.
We keep the assistant’s own explanation for picking them — clearer pricing, closer to the customer, more reviews, a page that answers the question. That reason is usually your fix, in AI’s own words.
The same questions, checked against the search results as well. Where the two overlap is where a competitor is strongest — and where a gap shows up in only one of them, that is worth knowing too.
04 · Perception
A report on what the assistants think you are, built from the answers we already stored. No extra questions — just a closer read of what came back.
Taken straight from the answers: how the assistants describe you when your name does come up. Sometimes it is not the description you have been paying for.
Which businesses AI mentions in the same breath as you. That is the shelf you are on, whether you picked it or not.
ChatGPT and Gemini side by side. When both tell the same story about you, that story is settled. When they disagree, there is room to change it.
05 · Topics
A topic is a job you want AI to send people to you for. Everything else on your report hangs off it.
“Emergency plumbing in Leeds.” “Project tools for small teams.” Plain words, the way a customer would put it.
Under every topic sit the questions a real customer would ask about it. Those are the questions we run, three times per assistant, every time you check.
You can see which topics you turn up in and which you are missing from entirely, so the work goes where the gap is instead of where it feels urgent.
06 · Fixes
A problem you cannot act on is just bad news. Every finding turns into something finished, ready to paste in.
Not “add a pricing page” — the actual paragraphs, or the actual code label, finished and ready to use. You are never handed homework and a hint.
The change most likely to move your score sits at the top. There is always one obvious next thing, so the list never becomes a pile.
Nothing is ever applied to your site automatically. We never touch your website — you, or whoever looks after it, makes the change.
When the change is live, run another check-up. Nothing is marked as fixed on trust.
Your Website Score moves as soon as you check again. AI Visibility takes longer — the assistants need time to notice a change, so it moves over later check-ups rather than the same afternoon.
07 · Evidence
A number you cannot check is a number you have to believe. None of ours are like that.
Click a score and read the answers it came from — question by question, assistant by assistant, exactly as they came back.
Every recommendation is tied to a finding, and every finding is tied to an answer we kept. If we cannot point at the reason, we do not make the suggestion.
Where one assistant agreed with itself, we say so as a count of its three tries. That is the whole measure of certainty we offer, and it is a real one.
We test ChatGPT and Gemini, and nothing else. Perplexity and Google’s AI Overviews are not checked, so no line on your report is a guess about them. When that changes, it will be a feature you can see running, not a claim on this page.
A free check-up takes a minute or two. No account needed to run it, and a free account saves the report.