What are the numbers right now?
| Metric | Baseline (July 4, 2026) | Latest (September 4, 2026) |
|---|---|---|
| Overall visibility Share of all AI answers (branded + unbranded prompts) that mention Lighthouse. | 47 | 37 −10 |
| Unbranded visibility Category questions ("best AEO tools") where engines name us without being asked about us. The metric that matters most — and the hardest to move. | 0 | 0 ±0 |
| Branded visibility Questions that name Lighthouse directly. | 100 | 100 ±0 |
| GEO score (avg, 20 pages) How citation-ready our own pages are: answer capsules, structure, original data, technical SEO. | 61 | 56 −5 |
| Our citations Times our pages were linked as sources across all engine responses in the run. | 16 | 88 +72 |
| Sentiment How answers characterize Lighthouse when we do appear (0–100). | 60 | 56 −4 |
The unbranded row is the honest one. A tool that measures AI visibility should be visible to AI — and at baseline, we weren't. We publish the zero because a scorecard that starts after the fixes proves nothing.
How do our citations compare with tracked competitors?
Citation counts from the same analysis runs — how often each brand's own pages were linked as sources across all engine responses. Competitor names and counts are our measurements, not their reported figures.
| Brand | Baseline | Latest |
|---|---|---|
| Lighthouse (us) | 16 | 88 +72 |
| Profound | 17 | — |
| Otterly | 16 | — |
| AthenaHQ | 10 | — |
| Evertune | 7 | — |
| Mention | — | 25 |
| Semrush | — | 24 |
| Hootsuite | — | 5 |
| Meltwater | — | 15 |
| Brandwatch | — | 17 |
| Talkwalker | — | 8 |
| Sprout Social | — | 8 |
Where do we show up, engine by engine?
Visibility per engine from the latest run (branded and unbranded prompts blended). Engines retrieve differently, so the spread is the point — a brand is never uniformly visible.
| Engine | Visibility |
|---|---|
| Claude Sonnet 5 | 48 |
| Claude Opus 5 | 48 |
| Gemini 3.7 Flash | 48 |
| Gemini 3.1 Pro | 48 |
| DeepSeek V3.2 | 48 |
| Qwen | 48 |
| GLM | 12 |
| Perplexity Sonar | 16 |
| Perplexity Sonar Pro | 24 |
| Google AI Overview | 28 |
| Llama 4 Maverick | 48 |
| Mistral Medium | 48 |
| Command A | 48 |
Are AI crawlers actually fetching our pages?
Visits to ahoylighthouse.com by AI crawlers over the last 30 days, logged by our own bot tracker. This is the first of the three gates made visible: no crawler visits, no citations.
| Crawler | Visits | Pages crawled | Active days |
|---|---|---|---|
| openai | 2379 | 716 | 61 |
| amazon | 690 | 448 | 28 |
| microsoft | 487 | 119 | 31 |
| apple | 219 | 116 | 9 |
| anthropic | 211 | 171 | 8 |
| perplexity | 184 | 149 | 13 |
| 148 | 133 | 4 | |
| meta | 93 | 54 | 30 |
| bytedance | 18 | 17 | 2 |
What are we doing about it?
The same sequence we recommend to customers, dated as we shipped it:
- July 3, 2026 — open the gates. Discovered our CDN was blocking GPTBot, ClaudeBot, and PerplexityBot; disabled the block, shipped an explicit welcome-policy robots.txt, sitemap, llms.txt, and agent-discovery endpoints (MCP server card, agent skills, markdown negotiation). Our site now scans as Level 5 "Agent-Native" on Cloudflare's public agent-readiness checker.
- July 4, 2026 — answer-shaped content. Shipped an answer hub of eight guides built the way we tell customers to build: answer-first blocks, question-shaped headings, specific values, visible dates. Also wired IndexNow and Google Search Console for faster indexing, and froze the baseline you see above.
- Next — off-site presence. Engines cite third-party sources heavily (Reddit, review sites, YouTube, listicles), so on-site work alone has a ceiling. This phase is slower and mostly human; the scorecard will show whether it registers.
The method behind each step is documented in the three-gates guide and the tracking method — this page is those guides applied to ourselves.