How we test AI visibility

Four steps, the same for every practice. No scraping of consumer chat apps. No invented ChatGPT sessions.

1. Crawl

ReplyScoreBot reads public pages you asked us to audit. We look for schema, name/address/phone consistency, question-form content, machine access, and trust signals. We obey robots.txt, cap depth, and identify the bot on /bot.

2. Query

We ask Anthropic and Perplexity Sonar the local questions patients actually type: best dentist near a town, NHS lists, nervous patients, Saturday emergencies. Each answer is stored so your report can show whether you were named, who was named instead, and the verbatim excerpt.

3. Benchmark

Site signals and surface results are scored against audited UK dental peers. A percentile is only published when the peer set is large enough to defend. Grades are letter bands on a 100-point scale, not a marketing label.

4. Fixes

The paid report ships prioritised fix cards, paste-ready JSON-LD, and an email you can forward to a web agency. Optionally, Reply Intelligence Group can implement the pack.

Why Anthropic and Perplexity?

We only test engines we can query through stable, official APIs, so every result in your report is reproducible and evidence-backed. We do not scrape consumer chat apps or simulate ChatGPT sessions, because those results cannot be verified and change day to day. Coverage of further engines (including ChatGPT and Gemini) will be added as stable interfaces allow, and the fixes in your report improve your visibility to all AI engines, because they read the same underlying signals: your site structure, schema, and reputation data.

See a sample report or run a free scan.