client names are under NDA until clients approve disclosure. every number below comes from our logged measurement panels - we show the raw logs on the call.
european crypto-law firm: from 1,000 invisible posts to a top-2 answer.
the firm came to us in June 2026 with a health score of 45/100. a YMYL niche (EU crypto licensing, MiCA), 1,004 blog posts - and zero of them in Google US top-3. the main competitor ranked ahead with 6.8 times less content. category recall across our AI panel: 0 of 5 runs. the engines’ complaint was concrete: anonymous content, no visible lawyers, no verifiable credentials.
what we did
- merged the blog-vs-money cannibalization: one content series was competing with the firm’s own product pages - 22 redirects, one hop, verified live
- built the machine-readable identity layer: llms.txt, LegalService schema, Person schema for all 8 lawyers with real credentials (PhD, bar license numbers, ORCID)
- turned anonymous articles into authored ones: bylines + “Reviewed by” with the founder’s credentials on money content
- claimed and cleaned the review surfaces the engines actually read
- re-measured the same 5-run panel every two weeks
the client’s own team shipped every change - 009 delivered hand-offs: exact code, exact wording, exact placement. that’s the model.
results
| value | what it is |
|---|---|
| 0/5 → 4/5 | category recall across the AI panel on day 17 of the program (measured run of July 18, 2026). the series since: 4/5, 4/5, 3/5 - we publish the series, not the best run. |
| #2 in Europe | where Perplexity placed the firm for EU/MiCA licensing in the August 15, 2026 run - citing the firm’s own service page as the source. |
| +21% | US organic traffic value vs the locked baseline (DataForSEO ETV, August 15, 2026), in a niche that cooled ~40% after the MiCA deadline passed. |
| 3.5× | brand mentions vs the nearest competitor in the measured corpus (August 2026). |
engine quotes from the panel logs: Claude, August 15 run - “among the most experienced for pure EU/MiCA CASP work”.
web3 & AI development agency: four engines, zero to named on all of them.
May 2026 audit: 27/100. AI visibility 5/100 - zero AI Overview citations against 628 for a competitor. worse: the brand name itself was contested - Google’s AI summary resolved it to an IEEE conference and a UN committee before the agency. dead URLs from an old site migration were still being cited as sources by the engines.
what we did
- rebuilt the redirect map to 41 rules, killing the zombie URLs engines kept citing
- shipped llms.txt, JSON-LD, and a brand fact audit: one impressive-sounding revenue figure turned out to be circular (our own materials quoting each other) - we removed it from every surface instead of publishing it
- restored and cleaned the directory profiles AI engines actually read, built the entity layer (Wikidata with press references)
- ran an article pipeline on demand-verified keywords: every topic backed by a live search-volume measurement, every draft through QA before the client saw it
- pitched real list-owners by playbook: named contacts, their asset first, follow-up schedule
results
| value | what it is |
|---|---|
| 27 → 53 | health score at the July 2, 2026 re-audit, seven weeks after the first audit. |
| 4 of 4 | AI engines that named the agency organically at least once within the first program month (July 28, 2026 panel; the composition shifts run to run - that’s the volatility we track). |
| top-10 in 2 weeks | two program articles entered Google US top-10 (positions 6.7 and 8.7) within two weeks of publishing (GSC, August 15, 2026). |
| source #1 | by August 2026, Perplexity builds its category answer on the agency’s own article as the first source; Gemini cited the domain 8 times in a single answer. |
positions and citations are single-run measurements from logged panels; we re-test on every review.