Frequently asked questions
Short, quotable answers about what TangentFlow measures, what it costs, and what we will not do.
What is TangentFlow?
TangentFlow measures whether AI assistants recommend your website. Paste any URL: we read the site, work out the market it serves, ask AI agents 30 live buyer questions about that market without naming any brand, and report who was recommended, how often, and the reasons the agents gave.
How is this different from SEO rank tracking?
Rank trackers watch a results page. We watch the answer. An assistant names three to five options and explains why; we sample that answer many times, count who appears, and record the stated reasons. Nothing in the report depends on a search ranking.
Do you guarantee that AI will mention my brand?
No, and nobody honestly can. We report what agents say, score whether your pages are readable to them, and give fixes that follow one public rulebook. Whether mentions move is measured afterwards and only reported when the numbers support it.
Can I pay to rank higher in your index?
No. Placement is never for sale at any price. The public index is never influenced by any paid service, and the same rubric applies to every site, including ours.
How many answers is a check based on?
30 live answers per check, plus 10 answers from a web-grounded surface used for the citation map. That is a sample, not a census: we report counts rather than a score, and call small movements what they are, not news.
Which AI models do you measure?
The public index runs on pinned model versions, currently a Claude Haiku 4.5 surface, with an OpenAI surface staged. Live checks add a web-grounded Claude surface with search enabled. Model versions are pinned rather than aliased, and every change bumps an epoch that is published on the methodology page.
What is an Agent Readiness score?
A 0 to 100 score from 17 objective checks on whether agents can read your site: crawler access, llms.txt, raw HTML content, structured data, pricing in text, identity, comparison content, docs, FAQ, freshness and entity presence. It measures readability, not popularity, and the two are never blended.
Do you write and publish content for me?
If you register a site, yes. Topics come from your own visibility data, drafts are written only from facts read off your pages, quality gates reject unverified numbers, and you approve every post unless you switch on auto-publish. Posts are delivered to your CMS by signed webhook or the WordPress REST API.
Is my data used to train models or shown to other customers?
No. Raw answers are archived so results can be re-judged and audited later, probe data never touches the public index, and nothing from one account is shown to another.
How much does it cost?
The live check is free for any site with no account. Monitoring and the content calendar are also free during early access; paid plans start at $199 per month when billing opens. See the pricing page for the full breakdown.
Still deciding?
Run a free check on your own site, or read the methodology for how every number is produced.