About PromptScore

What PromptScore is

PromptScore is a free diagnostic tool that measures how ready a website is to appear in answers from AI-powered search systems like ChatGPT, Perplexity, Claude, and Google AI Overviews. It runs 34 weighted checks across five categories, then returns a single score out of 100, a band label, a per-category breakdown, and a prioritised list of remediation actions you can act on the same day.

The tool is intentionally narrow. It does not try to replace a full SEO audit, a brand strategy, or a content programme. It answers one question well: where does this site stand, today, on the signals that AI retrieval and AI training systems actually use to decide whether to cite it.

Who built it

PromptScore is built and operated by Performance Peak Ltd, a digital and AI consultancy registered in England and Wales (registered office: 142 Nacton Road, Ipswich, Suffolk IP3 9JN). Performance Peak helps organisations adapt their websites, content, and operations to a search landscape that is rapidly being reshaped by large language models.

The PromptScore project was started by Lee Debnam, founder of Performance Peak. The motivation was simple: clients kept asking variants of the same question, “are we ready for AI search,” and there was no repeatable, transparent way to answer it. Existing tools either inspected one signal in isolation (robots.txt, schema, llms.txt) or returned vague opinion-based scores with no published methodology. We wanted both rigour and accountability. So we built PromptScore, published the full scoring methodology with every check and weight, and committed to versioning the model so changes are visible.

How it works

When you submit a URL, PromptScore fetches the homepage and a sample of inner pages, parses the HTML and structured data, inspects robots.txt and llms.txt, runs deterministic and composite checks against published specifications (Schema.org, llmstxt.org, the major AI crawler documentation), and grades a small number of subjective checks (homepage clarity, citation practice) using Claude at temperature zero with a locked, version-controlled prompt. All AI-graded results are cached against your content hash, so re-running an unchanged page returns the same score.

The full check list, weights, scoring bands, and AI prompt versions are published on the methodology page. If you disagree with a check or a weight, that is the right place to start the conversation; we update the model when the case is good.

Our principles

• Transparency over branding. Every check is public, every weight is justified, every AI prompt is versioned.
• Repeatability over novelty. The same input must produce the same score. AI-graded checks are temperature-zero and cached.
• Useful over comprehensive. 34 checks, ranked by impact. We would rather miss a marginal signal than drown the report in noise.
• Honest about limits. A score is a systematic estimate, not a guarantee of citation frequency. We say so, on the result page.

Get in touch

For corrections to the methodology, partnership enquiries, or commercial work with Performance Peak, email hello@performancepeak.co.uk. For removal of a specific scan result, email lee@performancepeak.co.uk with the scan URL and we will delete it from the database.