Research and original data

Everything here was measured rather than estimated, carries the date it was collected, and publishes enough method that somebody who disagrees can go and check.

We publish the measurements our own audit engine produces, including the ones about our competitors and the ones that are unflattering to us. Every study names its method, its sample and its collection date, and ships the raw data without a form. Take any of it and cite it.

11 August 2026

1048 of the most visited websites: who blocks the AI crawlers

28.3% of sites with a robots.txt block at least one AI crawler, against 1.5% that block Googlebot, so it is a decision rather than a locked door. 145 of them block a crawler that feeds live answers, which is usually not what they meant. 11.5% publish a real llms.txt.

Method
robots.txt, llms.txt and the homepage fetched for every domain, twice: once as a declared crawler and once as a browser. Rules evaluated per RFC 9309.
Data
Every domain and column published as CSV, both sample frames documented, CC BY 4.0.
Read the study

6 August 2026

98 GEO agencies, audited with our own engine

28 of 98 publish pages carrying no date at all and 22 name no author in the markup, which are the two signals they sell. Median score 87 of 100.

Method
36 checks against each agency's own homepage, concurrency 2, collected twice.
Data
Full CSV, sample frame and source pages published, CC BY 4.0.
Read the study

6 August 2026

11 AI visibility tools, priced and scored

Every price read from the vendor's own pricing page, and every vendor's own website scored on our engine. 11 of 11 fail freshness signals on their own site; the category averages 50 out of 100 on structured data.

Method
Prices read 2026-08-06. Scores from the same engine, concurrency 2, with the exact URL scored published beside every result.
Data
Every figure sourced to the page it was read from, including one vendor that publishes no price at all.
Read the study

5 August 2026

What GEO actually costs across the market

23 agencies with a published starting price, and only 4 of them published it about themselves. The rest come from third party comparisons, which is itself the finding.

Method
Every figure carries the page it was read on and the date. Read 2026-08-05.
Data
Sources linked per row. Nothing inferred: one agency was dropped for having no traceable figure.
Read the study

Why any of this is published

Two reasons, and the second one is the honest one. The first is that this category runs on assertions: prices nobody sources, percentages that trace back to another blog post quoting a third, and vendor claims with no method attached. Publishing measurement with its workings is the most useful thing a company with an audit engine can do in a market like that.

The second is that we sell the engine that produced these numbers. A study is a demonstration, and we would rather demonstrate on real data about real companies than in a case study we wrote about ourselves. That is a commercial motive and it is worth you knowing about while you read. It is also why every study here publishes the exact URL it measured and the thresholds it applied, so the demonstration can be run against us.

Run the same engine
on your own site

Free account, no card. Paste your URL and get a real, scored report of your AI and search visibility.

Measuring rankings in