We have: half your pages get the change, half stay untouched, and the difference between them is the answer. Install the plugin now — it sits idle, and the network starts the moment we reach 150 installs.
The threshold isn't marketing: below 150 sites the experiment would conclude nothing, and opening it anyway would mean selling a measurement that isn't one. The first 300 installs keep founding terms.
The first 150 accounts keep founding terms.
BotPass is a GEO platform that proves which changes to your site make AI assistants cite you. It doesn't estimate it or infer it: it runs a real experiment on your site. A randomly drawn half of your pages receives the change; the other half stays exactly as it is and serves as the control group. Several weeks later we compare how often the AI cites each group, and that difference — with its confidence interval — is the result.
What clears the bar enters a shared playbook and is applied automatically to every site with that page type. What doesn't is discarded, and we tell you about that too.
And it isn't your fault: until now there was no way to check.
More and more questions are answered inside the chat and never reach your site. The fight is no longer only about ranking: it's about being the source the AI uses to answer.
FAQ blocks, answering in the first forty words, question-shaped headings, structured data. It all sounds reasonable. Nobody has tested it against a control group.
With nothing to compare against, any lift can be attributed to any cause. That isn't measurement. It's a story that happens to fit.
All of GEO today is correlational. What was missing is the control group.
Five steps. Nothing enters the playbook without going through all five.
One change, from a closed catalogue. Never a free-text idea.
Pages are drawn at random into treated and control.
The plugin changes the treated half only. Always reversible.
We ask the AIs directly and log every crawler that returns.
The difference between groups, with its margin of error.
Half change, half don't. The gap between the two groups is the only thing that can tell you whether the change did anything.
If you change your pages and weeks later the AI cites you more, you don't know whether it was your change, a model update, a competitor who stopped publishing or the season. The control group lives through exactly the same things as the treated half except the change, so everything else cancels itself out.
It's the same principle used to approve a medicine, applied to web pages. It isn't sophistication for its own sake: it's the only known way to separate what works from what looks like it works.
And this is where the free part comes in.
An experiment needs around 150 pages in each arm for the difference between the two groups to separate from noise. A site with 300 pages of the same type barely makes it. A site with 60 doesn't come close — which is why most sites decide their GEO by folklore: not because nobody wants to measure, but because alone they can't.
The way out is to spread the same experiment across many sites at once. The pages add up quickly, but then the real problem appears: if they come from three or four sites they resemble each other too closely and the effective sample size collapses. What's needed isn't more pages, it's more different sites.
That's where the 150 installs come from. It isn't a rounded-up commercial target: it's the number below which the first experiment would conclude nothing, which is why it's on the front page instead of hidden.
Now each block is a different site. Inside every one of them, half the pages change. The experiment lives in the set, not in your site alone.
Ten minutes from your WordPress dashboard. Create a free account (Network mode) and paste the API key. No card.
It looks at which page types you have and how many, and touches nothing else.
The network goes live, your site joins the split and the first experiment begins.
Every experiment that closes, with its effect and interval. Apply it if you want.
What was tested, on which page type, what effect came out and with what confidence interval. The whole record, not a headline.
Whatever clears the bar you can apply from the plugin, and revert just as fast. It doesn't touch your template or your styles.
First 150 accounts: whatever the Network's final model turns out to be, for you it never gets worse.
First 150: we start with the page type that has the most founding sites in it. Install now and your type goes first.
First 150: during the first cohort we read the emails ourselves. That's the real reason it's 300 and not 3,000.
If your site turns out to have the volume, you'll be able to run your own control group inside your own site. We'll tell you, not a salesperson.
This is a free WordPress plugin, not a fourteen-day trial in disguise. So the list of what there isn't is short and it's literal.
What we do ask is that you leave it on. An experiment takes weeks to close, and a site that comes in and drops out after a fortnight adds nothing to the measurement — not for you and not for the rest of the network.
All three are legitimate. They answer different questions, and only one answers why.
| BotPass | AI visibility tool | GEO agency | |
|---|---|---|---|
| Tells you what happened | Yes | Yes | Yes |
| Tells you what caused it | Yes — control group | No | No |
| Changes your pages for you | Yes, automatically | No | Yes, by hand |
| Every change reversible and logged | Yes | n/a | Depends |
| Gets better from other sites' results | Yes — shared playbook | No | No |
| Will tell you a change did nothing | Yes, in writing | n/a | Rarely |
If you already have a visibility tool, keep it: we can read its data as a measurement source. This doesn't come to replace it.
We're a GEO and digital marketing agency based in Barcelona. We've been doing this for clients for years, and BotPass comes out of exactly the frustration you have: recommending changes without being able to prove which ones work. We built it first so we could answer that question in our own work.
That's why the plugin is free, and why the 150 threshold is written on the front page instead of hidden: if the network doesn't get there, we'll say so.
Free, no account, no card and no call. It sits idle until the network goes live.
BotPass.zip with the button aboveWhat it does the moment it activates: it checks which page types your site has, how many of each, and whether it can read the origin logs. That's all. It doesn't modify a single page or add weight to your site until the network starts.
Leave us the domain and we'll let you know when the network goes live. One email, nothing else.
By submitting you accept our privacy policy. No payment method is requested or stored.
Nobody controls what a model answers. Anyone who guarantees it is lying, and you'll find out in six months.
Until the network reaches 150 the plugin is idle, and after that an AI crawler takes weeks to come back and re-read a site. If you're expecting a dashboard full of numbers on day one, this isn't your product.
Plenty of it will come back flat. We'll tell you in the same detail as if it had worked, because knowing what doesn't work is part of the product too.
We sell certainty, not promises. And today what we have is the method.
Optimising your site so that the models answering questions — ChatGPT, Perplexity, Claude, Gemini, Google's AI overviews — use it and cite it as a source. SEO optimises for appearing in a list of links; GEO optimises for appearing inside an answer that often produces no click at all. They coexist, but they're measured differently, which is why one's recipes don't automatically work for the other.
Nothing visible. On activation it checks which page types you have, how many of each, and whether it can read your server logs, then it sits idle. It doesn't modify any page, doesn't change your content and adds no weight to the site. When the network reaches 150 installs we'll tell you before anything starts running.
We say so publicly and no experiment opens. The plugin won't have touched a single page of your site and it uninstalls from the WordPress dashboard like any other. Nobody paid anything, so there's nothing to refund and nothing to get out of.
Those tools measure from the outside: they query the models and report what came back. They tell you what happened, and they do it well. They can't tell you why, because nothing on your site changed on purpose. We change a randomly drawn half of your pages and compare against the untouched half; that comparison is what licenses talking about causes. Keep your tool — we can read its data as a source.
Free. No expiry date and no payment method of any kind, neither at install nor when the network goes live. BotPass is monetised when a site with its own volume wants its own experiments inside its own website. And the first 150 accounts keep founding terms: whatever happens to the model, for them it doesn't get worse.
The sites in the Network are what make it possible for an experiment to reach a meaningful size, and in exchange they get the result. When a site has enough volume to experiment on its own, it wants to do that on its own site with its own hypotheses — and that part is paid. We don't sell anyone's data: what's shared is the aggregate effect of each change, never your content, your traffic or your domain.
It applies variations to the pages that fall into the treated arm, and only to those: never your whole site. It doesn't touch your template or your styles, it sits alongside Yoast and RankMath, and any change reverts in one click leaving a record of what was applied, where and when. Changes that alter visible text wait for your approval.
No. Exactly the same content is served to people and to machines, and there's a parity report, by URL and date, documenting it for your SEO lead and your legal team. Serving different content to crawlers is cloaking; this isn't.
No minimum. The Network exists precisely for the sites that don't get there on their own. If it turns out you do have the volume — more than 300 pages of the same type — you'll also be able to run your own experiments inside your own site, which is a better conversation for you. The plugin will tell you on install.
Today the plugin only exists for WordPress, because page-level randomisation and reading the origin logs need that access. Leave us your domain in the notification form and we'll write if your platform comes in.
Install them and write to us, because your case has a technical wrinkle: forty sites managed the same way resemble each other, and for the experiment they count for considerably less than forty independent ones. It's still very interesting, it just gets organised differently.
Performance Media, a GEO and digital marketing agency based in Barcelona. BotPass is our product, built in-house and operated by San Mateo Capital, S.L. You can write to hello@botpass.io — during the first cohort we answer it ourselves.
Ten minutes, free, no account. The plugin sits idle and the network starts when we reach 150 installs — right now 150 to go.