One cockpit over four planes — SEO, AEO, GEO and AI-Visibility. At-a-glance verdict, what to do next, and a launchpad into ~50 deep tools. Real data, honest tiers.
live data · 21 Jun 2026
anthropic.com· last scanned 2026-06-21
The Verdictihow the score is built · honestOverall score 56 = the average of the three on-page planes you control (SEO 62 · AEO 38 · GEO 69). AI-Visibility 37% is the measured reality those planes are supposed to earn — measured, not modelled. Every figure is live from anthropic.com — four planes, 29 tools, a 475-page crawl, each reviewed against its source. SEOptimer grades the same site B+ but stamps GEO an F; we publish the math behind all four planes.Do → close the gap with the cross-plane schema + freshness fixes; we re-run the exact prompts to prove the lift.
Built better than you're seen.
You ship the models that answer every buying question — and they barely name you. Known (GEO 69) and findable (SEO 62), but not built to be quoted (AEO 38) — so your AI-Visibility cites you just 37%: 5 of 22 AI Overviews, 0 of 16 ChatGPT buying questions. Not a capability gap — a visibility gap. Every figure below is live from anthropic.com.
◆
The four scoring planes
Each plane is a labeled 10-signal radar — every spoke is a ranking factor the deep tools measured, shaded in the plane's own colour (light = low, deep = high). Switch to ⊕ ALL to overlay the planes on shared signals and see which factors move several at once. Nothing here is a dead end.
◆
The intelligence layer
The four intelligence planes — Competitors, Technical, Citations and the Action Center — each a labeled 10-signal radar. Competitors is a directional parity composite (you vs the category leader) and Action is deployment-readiness (0 = not yet shipped); Technical & Citations are measured.
The whole site — every page downloaded & scored (475 URLs)
Every page as a dot — ranked by authority
474 of 475 URLs · X = page rank · click any dot for its full breakdown
imeasured · page-rank compositeEach dot is one real URL, placed by its page rank — a 0–100 composite of citation authority, on-page rank-readiness & content depth. Per-URL search volume & traffic aren't measured (site-level only). No page clears 80 — there's no standout URL.Do → a handful of product/careers pages top the chart; fix the template, not 475 pages.
hover a dot for the page & why · click for the full breakdown
X = page rank (citations × readiness × depth, locked) · pick any Y factor above to plot against it · no page clears 80 — the site has no standout URL
0% of pages declare a date — exactly 1 of 474 carries datePublished/dateModified. Recency-biased engines drop undated sources first.
Only 2% carry schema (11 of 474) — and those 11 product/careers pages are the best-scoring pages, proving the lever works site-wide.
News (231) + Research (143) = 79% of all URLs — the biggest content investment is exactly where 0% dates & 0% schema apply.
The competitive reality — capability decoupled from visibility
Company & product share of voice — who gets mentioned, who leads
86 AI answers · companies and products · breadth = mention share · leadership = rank when cited
imeasured · 86 AI answersBreadth = how often a brand is named in AI answers; leadership = its average rank in the answer (1 = named first). You're a strong #2 on breadth but usually cited 2nd–3rd.Do → push presence before chasing first-choice — get into the room.
Head-to-head — you vs any competitor, every dimension
pick a lab · organic · overall reach · AI presence · brand sentiment
The link constellation — your backlink network in 3D
Not a decoration — a map of how the web actually links to you. The core is anthropic.com; the inner ring is the real pages that earn the links (your careers page and homepage pull the most), sized by link volume; the outer nodes are the referring domains, sized by Authority Score, coloured by category, each wired to the page it points at. Hit vs OpenAI to see the authority gap as a picture — their web is 3.4× denser.
scroll to zoom · drag to rotate · hover a node for its link · size = domain & page value
A working prototype of the live viz v1
Self-contained and live in your browser (no dependencies). Real top-referrer domains + real linked pages drive the structure; the surrounding cloud is a profile-accurate scatter of the full 123,340-domain network (sized by the real Authority-Score pyramid, coloured by the real category mix). The full Semrush backlink suite — Authority-Score drivers, the 160K-domain outreach gap, anchor & toxicity audit — lives in the Backlinks tool.
04
Citations and mentions
Where your citations & mentions come from
each AI engine · node size = referral weight · tendril width = mention rate
measured · 5 engines + Reddit
ChatGPT is huge but thin — ~77% of AI referral weight, cites you 27%. Reddit out-cites every lab — 13 of 22 AI Overviews. The traffic and the citations point at different doors.
Organic footprint
live · United States · 12-mo direction (Semrush)
net −35K rankings
+51.8K new rankings − 86.8K lost = net −35K. The footprint is eroding even as off-page authority (AS 69) rises — capability decoupled from discoverability.
Top keyword answers — who gets cited, engine by engine
The empty ChatGPT column is the "0/16 commercial" story — a vertical band of misses where the traffic lives. You're cited everywhere for your own inventions (constitutional AI, MCP); for the commercial terms buyers actually ask, you're absent and the engine quotes Reddit.
05
Top 10 fixes — click any to see how it moves each plane
One fix, four planes › The leverage here is that the same repairs move multiple planes at once. Emitting entity schema lifts SEO E-E-A-T, GEO entity authority, AEO extraction, and AI-Visibility citations — simultaneously. Deploy, then we re-run the exact prompts to prove the lift.
Projected readiness imodel_estimateModeled readiness after the cross-plane fixes deploy. We then re-run the exact prompt set (163 replays) to prove the real lift — not just claim it.Do → deploy entity schema first (moves 4 planes), then re-verify.
Per-plane lift is modeled (a projection), then re-verified by re-running the exact prompt set — the Action Center loop. The 1–10 order is by leverage across the four planes; click any fix to open its per-plane impact.
▦
Open any suite — the full toolset behind the score