Scores you can trust. Gaps we admit.
Every number comes from real crawl data, live AI probes, or verified APIs — never from an LLM guess. When we can't measure something, we tell you and exclude it from your score.
Core principle
Deterministic numbers. AI-written narratives.
The score NUMBER you see — 74, 38, 91 — is always the output of deterministic rule-evaluation against real measured data. We crawl your site, call the Google PageSpeed Insights API, run live probes on ChatGPT and Perplexity, and evaluate each signal against a fixed rubric. There is no LLM guessing your score.
The narrative text — the “what to fix” explanations, the issue descriptions, the recommended actions — those are written by AI from the evidence we collected. That is where language models add value: synthesizing structured data into human-readable recommendations, not producing the number.
Score = deterministic rule-evaluation
Same site, same rules, same score. Reproducible by construction.
Narrative text = AI-written from evidence
LLMs synthesize findings into plain-English actions. They never produce the number.
Weighted composite of 4 pillars
SEO 30% · AI Search 25% · Social 25% · Design 20%
Confidence layer
When we can't fully measure, we say so.
Some sites block our crawler (Cloudflare-protected, login-gated, or otherwise inaccessible). When that happens, we don't score those components at all — and we renormalize the score over what we could actually measure.
The result is an honest partial score with a coverage indicator (High / Medium / Low) that tells you how much of your brand we were able to reach. A site scored at Medium coverage got a real score — just over fewer dimensions than a fully accessible site.
High coverage
All four pillars measured. Score is fully comparable across audits.
Medium coverage
Most pillars measured. One or two components excluded due to access limitations. Score renormalized over what was reached.
Low coverage
Significant portions unreachable. Score shown with explicit “couldn't fully access this site” disclosure. Never scored as if the missing data were a performance failure.
Live probes
Not cached data. Live measurements.
AI Search scores come from live queries run against real AI platforms at audit time — not from a cached database of “what AI said last month.” Each audit fires 17 real prompts across ChatGPT, Gemini, Claude, and Perplexity and measures whether your brand appears, where it appears, and how it's described.
This means your AEO score reflects today's reality. If you publish a new piece of content and it gets picked up by AI training, re-running the audit will show it. The data is never more than 24 hours old.
ChatGPT
GPT-4o · brand + category queries
Gemini
Gemini 1.5 Flash · local + expert queries
Claude
claude-3 · comparison + recommendation queries
Perplexity
Sonar · real-time web-augmented queries
Composite score
How the four pillars combine
The overall brand score is a weighted average of four pillar scores. Weights reflect the relative impact each domain has on AI-era customer acquisition.
| Pillar | Weight | What it captures |
|---|---|---|
| Website / SEO | 30% | Crawlability, on-page optimization, Core Web Vitals, structured data, content depth, off-page authority (DataForSEO) |
| AI Search (AEO) | 25% | Citation rate, AI share of voice, platform coverage (ChatGPT/Gemini/Claude/Perplexity), position in answer, entity clarity, structured data |
| Social | 25% | Platform presence, sharing infrastructure, profile completeness, posting recency |
| Design | 20% | Mobile readiness, accessibility, trust signals, brand consistency, page completeness |
Weights source: src/lib/constants.ts — SCORE_WEIGHTS
Pillar 1 — Website / SEO
Technical and content signals that search engines measure
Composite weight
30%
The SEO score measures how discoverable your site is to both traditional search engines and the AI crawlers that power answers on ChatGPT, Perplexity, and Google AI Overviews. It combines six sub-components across technical, on-page, performance, content, and off-page dimensions. All six are measured deterministically from crawl data, Google PageSpeed Insights, and the DataForSEO backlink index — no LLM estimates.
| Sub-component | Weight | What we measure | Data source |
|---|---|---|---|
| Technical Crawlability | 22% | robots.txt rules, sitemap presence, canonical tags, redirect chains, HTTPS, and AI-bot access (GPTBot, PerplexityBot, ClaudeBot). A site that blocks AI crawlers loses discoverability across every AI-search platform. | Own crawler — fetches robots.txt, sitemap.xml, HEAD responses |
| On-Page Optimization | 22% | Title tag length and keyword presence, meta description, heading hierarchy (H1/H2/H3), internal linking density, keyword usage in body copy, and image alt text coverage. | Own crawler — multi-page HTML parse |
| Core Web Vitals | 18% | Google's 2026 performance thresholds: LCP under 2.5 s, INP under 200 ms, CLS under 0.1 — measured from real-user data via the PageSpeed Insights API. Mobile and desktop scores reported separately using Lighthouse weights (TBT 30%, LCP 25%, CLS 25%, FCP 10%, SI 10%). | Google PageSpeed Insights API (CrUX + lab data) |
| Structured Data | 14% | JSON-LD schema presence and completeness: Organization, LocalBusiness, Product, FAQPage, Review, BreadcrumbList. Validates against the Schema.org spec — presence plus validity. FAQPage schema is a direct input to AI-answer citation probability. | Own crawler — JSON-LD extraction + schema.org validation |
| Content Depth | 14% | Word count per page, topic coverage across the site, FAQ content presence, and publication freshness (sitemap last-modified, HTTP Last-Modified headers). Thin or stale content is a leading predictor of low AI-answer citation rates. | Own crawler — multi-page text extraction + sitemap.xml |
| Off-Page Authority | 10% | Referring domains count, domain authority score, dofollow-link ratio, and SERP keyword visibility (top-10 and top-3 rankings, SERP feature ownership). Industry consensus places off-page signals at 40–50% of ranking power; Lume factors them at 10% of the SEO composite. | DataForSEO API — backlink index + keyword rank data |
Weights source: src/lib/analysis/seo-analyzer-v2/weights-v1.ts — SEO_V2_WEIGHTS
Pillar 2 — AI Search (AEO)
How AI engines answer questions about your business
Composite weight
25%
The AI Search score measures four outcomes from live probes across ChatGPT, Gemini, Claude, and Perplexity: Citation Rate (how often your brand is named in answers), AI Share of Voice (your mentions vs. competitors'), Platform Coverage (which AI platforms cite you at all), and Position in Answer (where you appear when cited). These are direct measurements from 17 real prompts per audit run — not proxies or estimates. The score also factors in entity clarity and content freshness, which are the upstream inputs that predict citation outcomes.
| Sub-component | Weight | What we measure | Data source |
|---|---|---|---|
| Structured Data for Citation | 30% | JSON-LD richness optimized for AI citation: FAQPage, HowTo, Organization with sameAs links, LocalBusiness with opening hours. AI engines preferentially cite structured, citable sources. FAQPage schema directly increases the probability of appearing in AI-generated answers. | Own crawler — JSON-LD extraction + schema.org validation |
| Entity & Knowledge Graph Readiness | 30% | How clearly AI engines can identify the business as a distinct named entity: consistent NAP (name/address/phone) across the web, sameAs schema linking to Wikidata/LinkedIn/Crunchbase, brand mentions on authoritative domains, and Reddit/community forum coverage. Entity clarity is the upstream prerequisite for AI citation — approximately 92% of AI Overview citations come from domains Google can unambiguously identify. | Wikipedia API, own crawler for NAP + sameAs consistency, community signal scraper |
| AI Citation Rate, Share of Voice & Platform Coverage | 20% | Four measured outcomes from live AI-search probes across ChatGPT, Gemini, Claude, and Perplexity: (1) Citation Rate — share of 17 tested prompts where the brand is named or linked; (2) AI Share of Voice — brand mentions as a fraction of all brand mentions in those responses, weighted by answer prominence; (3) Platform Coverage — how many of the 4 platforms cite the brand at all; (4) Position in Answer — average rank when the brand appears in a ranked list. These are direct outcome measurements, not proxies. | Live multi-prompt probes — 17 prompts across ChatGPT / Gemini / Claude / Perplexity; citation_rate and position stored per prompt run |
| Content Freshness | 10% | Age of the most recent sitemap entries, last-modified headers, and blog/news publication frequency. AI engines favor authoritative sources that show recent activity — stale content signals a less reliable source. | Own crawler — sitemap.xml + HTTP Last-Modified headers |
| Sentiment in AI Answers | 10% | When AI engines mention the brand in their answers, is the context positive (recommended, leading, trusted), neutral (mentioned), or negative (cautionary)? A business cited positively should score higher than one cited with caveats. Scored from the retrieved AI response text for each prompt. | Live AI-search probes — sentiment classification on citation context |
Weights source: src/lib/analysis/aeo-analyzer-v2/types.ts — AEO_V2_WEIGHTS
Pillar 4 — Design
Mobile readiness, accessibility, and trust
Composite weight
20%
The Design score evaluates the quality signals that directly affect both user trust and AI-engine authority scoring. Poor mobile experience, inaccessible markup, and missing trust signals each reduce the probability that an AI engine will cite your business as authoritative. All five sub-components are measured deterministically.
| Sub-component | Weight | What we measure | Data source |
|---|---|---|---|
| Mobile Readiness | 30% | Mobile usability score, viewport meta tag, touch-target sizing, font-size legibility on mobile, and absence of horizontal scroll. Measured via Google PageSpeed's mobile audit. | Google PageSpeed Insights API (mobile audit) + own crawler |
| Accessibility | 25% | Google Lighthouse accessibility score: image alt text coverage, heading hierarchy, color contrast (WCAG AA), form label associations, ARIA usage. Accessibility gaps also hurt AI crawler comprehension. | Google PageSpeed Insights API (accessibility category) |
| Trust & Credibility | 20% | HTTPS, visible contact information, privacy policy presence, trust badges, review widget presence, SSL certificate validity. These signals directly affect both user trust and AI-engine authority scoring. | Own crawler — element detection + SSL check |
| Brand Consistency | 15% | Logo present in header, consistent color usage across pages, professional typography, absence of broken images. Brand consistency is a proxy for how seriously a business treats its online presence. | Own crawler — multi-page visual element analysis |
| Page Completeness | 10% | Presence of key pages: About, Contact, Services/Products, and a working 404 page. Missing standard pages signal incomplete web presence to both users and AI engines. | Own crawler — URL existence checks |
Weights source: src/lib/analysis/design-analyzer-v2/weights-v1.ts — DESIGN_V1_WEIGHTS
Measured vs. estimated
What we measure directly, and what we estimate
Measured (deterministic)
- Own crawler: technical crawlability, on-page, structured data (SEO + AEO), content depth, social link presence, sharing infrastructure, trust signals, brand consistency, page completeness, entity/NAP consistency
- Google PageSpeed Insights API: Core Web Vitals (LCP/INP/CLS, 2026 thresholds), mobile usability, accessibility scores
- DataForSEO API: referring domains, domain authority score, SERP keyword rankings and features
- Live AI-search probes: citation rate, AI share of voice, platform coverage, position in answer, answer sentiment (ChatGPT, Gemini, Claude, Perplexity — 17 prompts per run)
- Wikipedia API + community signals: entity recognition, brand mentions on authoritative sources
Estimated (improving)
- Social profile completeness — estimated from website signals until Phase 2 social scraper is live. Affects 20% of the Social score.
- Social posting recency — estimated until social scraper can read platform APIs directly. Affects 20% of the Social score.
Sub-components that are currently estimated are labeled in the tables above.
See your score — and exactly how we got there.
Free audit in 60 seconds. No credit card. Full sub-component breakdown included.
Pillar 3 — Social
Platform presence and sharing infrastructure
Composite weight
25%
The Social score measures whether your business has a defensible social presence — claimed profiles, proper Open Graph tags, and evidence of active posting. 60% of this score is deterministically measured from your website and crawl data right now. The remaining 40% (profile completeness and posting recency) will shift from estimated to measured when our social scraper completes.
Weights source:
src/lib/analysis/social-analyzer-v2/weights-v1.ts — SOCIAL_V2_WEIGHTS