5 pillars. 100 total points. 4 tiers. The same scorecard powers the free FACTS Score and the work we do for clients — there is no "secret" tier.
Every audited site is scored across 5 pillars. Each pillar is worth a maximum of 20 points. The total (100 points) maps to one of 4 tiers below.
Within each pillar, the engine evaluates concrete, binary signals — schema present or absent, llms.txt valid or not, lastmod dates varied or always-today. No subjective grading. Disqualifiers (broken canonicals, malformed JSON-LD, etc.) deduct points AFTER the pillar total.
Every signal we check, every penalty we apply, and every tier you can land in is spelled out on this page. We run the same audit on our own website — our score lives at /facts/score. It's whatever the tool gave us this month, not a number we picked.
Every FACTS pillar moves you through the same path an AI agent walks before it names a business to a customer. Miss a step and you drop out of the answer.
Where your score lands tells you whether agents can see you, read you, or recommend you.
| Tier | Score range | What it means |
|---|---|---|
| Invisible | 0–40 | Functionally invisible to AI agents. |
| Developing | 41–60 | Partial signals — agents can find you, but not understand you. |
| Competitive | 61–80 | Ahead of most. Vulnerable to the next competitor that figures it out. |
| Dominant | 81–100 | Top tier. Maintenance, not catch-up, is what matters now. |
This is where your site scores today — not the plan you buy. Every plan moves you up these tiers; the tier just tells you how far you have to go.
Click any pillar to read the deep dive. Each links to its own page with the full signal list and common findings.
Can a machine tell who you are, what you do, and where you serve — without guessing?
If it can't, you never make the shortlist the customer actually calls.
Fact Architecture is the Schema.org/JSON-LD layer on every page that describes your business as structured facts — Organization, LocalBusiness, Service, FAQPage, Article, BreadcrumbList. AI agents extract these facts before they ever read your prose. Without them, you're ambiguous — and ambiguous businesses get skipped.
When someone asks AI a question you can answer, can it lift a clean answer straight off your page?
If your answer is buried in marketing fluff, AI quotes the competitor who made theirs easy to grab.
Answer Clarity scores how easily an AI agent — or a hurried customer — can lift a clean, self-contained answer out of your page. Semantic HTML, clear heading hierarchy, FAQ blocks, and short answer paragraphs win citations and conversions. Walls of marketing prose lose to competitors who structured their answers.
Can AI agents actually reach and read your site — or do they bounce off and move on?
Sites agents can't access don't get cited — you're invisible to the tools your customers now ask first.
Connection Surface is the set of machine-readable surfaces that let agents index, talk to, and transact with your site without scraping it. A well-formed llms.txt, a fresh sitemap.xml, RSS feeds, and clean public endpoints are the foundation. On top of that, three 2026 cross-platform standards have emerged that agents from OpenAI, Google, Microsoft, and Anthropic are aligning around: A2A AgentCard for agent-to-agent discovery, Universal Commerce Protocol (UCP) for AI-driven checkout, and WebMCP for runtime tool registration in Chrome. Sites that implement these are visible to the agentic web; sites that don't are invisible to it.
Can AI verify you're worth recommending — your reviews, ratings, and proof in a form it can actually read?
When two businesses answer equally well, AI recommends the one whose reputation it can see.
Trust Signals are the reputation cues an agent can extract directly from your site: AggregateRating schema backed by a real rating and review count, an Organization.sameAs array linking your social and industry-directory profiles, and case studies with named clients and outcome metrics expressed as numbers, not adjectives. Agents weight self-declared authority heavily — but only when it's machine-readable.
Can AI tell your information is current — not a business that may have closed years ago?
Agents skip stale sources for fresher ones, even weaker ones. Freshness keeps you in the running.
Signal Freshness is the set of cues that tell an agent your content is current: last-modified dates that vary realistically, content velocity (new posts, updates), refreshed pricing, and recent reviews. Agents prefer fresh answers; a stale site gets passed over for a fresher competitor even when the fresher one is less authoritative.
Run the methodology against your site. Free, 60 seconds, no credit card.
Free FACTS Score