How it works
Ask, measure, fix, prove
Buyers ask assistants before they ask you. This finds out what those answers say about your brand, tells you what to change, and shows you next week whether it landed.
We work out what your buyers ask
Onboarding reads your site and drafts the questions somebody would actually type into an assistant when they are ready to buy, in your category and your country. You edit them, and they become the set we track.
Questions that name your own brand are kept separate. An engine repeating a name you handed it proves nothing about whether it recommends you, so those are scored as brand recall and never mixed into visibility.
Every engine answers, and we keep the receipts
Each question goes to the engines on your plan. We record whether the answer named you, linked you, which rivals it named instead, and which pages it read to decide.
Google AI Overviews is included, which most tools cannot do: it has no API and only exists on the results page. When Google declines to render one, that slot is left out of the score rather than counted as a miss.
You get instructions, not a score
A number tells you that you have a problem. The console files what to do about it: the page returning an error, the search you are one edit away from winning, the article that does not exist yet, the review site answering for a rival.
Each one states its evidence. Not "improve this keyword" but "you sit at 12 for this, which is page two, and 4,820 impressions are already landing on it".
Next week shows whether it worked
Mark a fix applied and it is stamped on the trend. The following run either moves or it does not, and you can see which change it followed.
This is the part a one-off audit cannot do. A report is a photograph; the thing you need is the difference between two of them.
Running alongside it
Keyword and page tracking
Your Google position for searches you choose, and a weekly readiness score for pages you nominate, so an edit and the number it moved carry the same date.
The citation gap
Sources an engine read to recommend a rival and never to name you. Getting onto those puts you in the answer without outranking anybody.
Backlinks, yours and theirs
Referring domains and authority for your brand and every competitor, refreshed weekly. Authority still decides which pages an assistant reaches for, so it belongs next to the visibility numbers.
The full competitor matrix
Every tracked question against every rival, up to twenty of them. You can see which questions each one owns, and which sources put them in the answer.
Drafts for what is missing
Where the fix is an article that does not exist yet, the GEO Writer drafts it against the question being asked and the sources those answers already cite.
Search Console, reconciled
Your true average position at any depth, sitting beside our own reading, so a keyword ranking past page five still tells you exactly where it is and what it is earning.
A report you can forward
Every run exports as a PDF with the charts behind it, and it arrives by email each week. On Agency it carries your branding instead of ours.
API, MCP and the browser
Pull every score into your own reporting over the API, plug a workspace into Claude or Cursor through MCP, or score any page you are looking at from the free Chrome extension.
Backlinks, the full matrix and Search Console are on Growth and Agency. The API and MCP server are on Agency. The Chrome extension is free for anybody. Compare plans
Before you start
How long before I see the numbers move?
Your first full reading lands on day one, so you know exactly where you stand before you change a thing. After that, schema and metadata fixes are the quickest to show up, often within a week or two of going live. Bigger pieces like a new article build over a month or so as the engines re-read your pages. Every run is stored, so you watch the movement happen instead of guessing at it.
What is different about week four compared with week one?
Week one gives you the picture: where you stand, who is being named instead of you, and the ranked list of what to fix first. Week four gives you the line. By then you can see which of your changes moved something, which engines shifted together, and where you are gaining on a competitor. Both are useful, and the second is what a one-off audit can never give you.
Do I have to make the fixes myself?
You or your developer apply them, and most arrive with the code already written. Recommendations come as JSON-LD, meta tags or HTML you can paste straight in, with the page each one belongs on, so a lot of them are a two-minute job. Nothing is installed on your site and no access is needed, which means there is no integration work before you start and nothing to unpick if you stop.
What happens when an engine gives no answer?
That slot is left out of the score instead of counted against you. Google does not render an AI Overview for every search, and an assistant sometimes declines a question. Scoring those as a miss would punish you for something you cannot influence, and would make your chart move on Google's behaviour instead of yours. The run records that the slot was empty so you can see it happened.
Can I change the questions after setup?
Yes, any time, and removing one frees its slot straight away. Tracked questions are a standing capacity rather than a monthly allowance, so nothing accrues and nothing is wasted. The console also shows you the questions Google puts next to yours, so the set usually gets sharper after a few weeks of running.
Does it work if my site is new or has no traffic?
Yes, and a new site is where it pays off quickest. You get the questions your category is already being asked, who is getting named on them today, and the exact sources those answers are built from. That source list is the shortcut: getting onto pages the engines already trust puts you in the answer without having to outrank anybody first, which is a far shorter route than waiting to build authority.
Easier to look at than to read about
The preview is a full workspace ten weeks in, with nothing to sign up for.