Reports · Original research

Numbers, not opinions.

Research we run ourselves, published in full. No form, no gate — method and sample stated on every page.

Almost everything written about AI search is a guess. The field is two years old, the engines publish little about how they choose what to quote, and most of what fills the gap is opinion wearing the costume of expertise — confident, unfalsifiable, and sourced to nothing.

This section is our attempt to do the opposite. Every report here starts from data we gathered ourselves: crawls of real sites, aggregate results from our own free tools, Search Console exports shared with permission. We state the sample. We state the period. We publish the method next to the finding, so you can decide whether the finding holds rather than whether we seem trustworthy.

All of it is free and none of it is gated. That is not generosity — it is the strategy. A number locked behind an email form cannot be quoted, linked, or cited, and a finding nobody can cite may as well not exist. We would rather be the source than the sales funnel.

The first report is still in the field. In the meantime, the audit below is running on real sites — and its results are what the research is built from.

Run the free audit →
How the research works

Where the data comes from.

Most of it comes from our own tools. The AI Readiness Audit is free, needs no sign-up, and checks a real site against the things an answer engine actually looks for — structured data, crawlability, entity clarity, evidence of expertise. Every run adds an anonymous data point to an aggregate picture of what the web is getting right and wrong, and that picture is what most of these reports are built from.

The rest comes from work. The SEO Recovery Engine and the Intent Cannibalization Autopilot both run on real Search Console data, in the browser, and the patterns they surface across many sites are worth writing down. Where we use a client's data, we use it with permission and we report it in aggregate — never a named site, never a figure that identifies one.

And where somebody else has already done the work properly, we cite them and link out rather than re-run it badly. Original research is expensive. Pretending to have done it is cheap, and it is the reason so much of this field is noise.

See the tools the data comes from →
Questions

Common questions about the reports.

What is a Troiana report?
A piece of original data research. Every report starts from data we gathered ourselves — usually from running our own tools across real sites — rather than from a survey of opinion or a summary of somebody else's study. Each one states its sample size, the period it covers, and the method used to gather it, so you can judge the finding rather than take it on faith.
Do I have to give an email address to read one?
No. Every report is free, ungated, and published in full as a web page. There is no form, no sign-up, and no drip sequence. A PDF copy is offered where one exists, but it is a convenience — the entire report is on the page.
Where does the data come from?
Mostly from our own tools. The free AI Readiness Audit checks how ready a site is to be understood and cited by AI search engines, and every run adds to an anonymous, aggregate picture of what real sites get right and wrong. Reports also draw on public data — Search Console exports shared by clients with permission, crawls of public pages, and published research, always cited.
Can I quote or republish a finding?
Yes, and you are encouraged to. Quote any number, chart or claim with a link back to the report it came from. That is what the research is for — the reports are structured, marked up and licensed to be cited, by people and by answer engines alike.
How is a report different from an insight or a book?
An insight explains something that is already known. A report establishes something that was not — it is the one of the three that produces new data. A book is long-form writing collected into a PDF. If a report is the finding, an insight is the explanation, and a book is the argument at length.
How often do you publish?
When the data supports a finding, and not before. Research on a thin sample is worse than no research, so a report ships when the numbers are strong enough to defend rather than on a content calendar.