Capture conditions
All captures ran July 11, 2026. Full transcripts and screenshots are archived. Four conditions are disclosed because each one limits what the evidence can bear.
The logged-in contamination, and the control for it
The first ChatGPT runs came from a logged-in account, and the model itself volunteered that it knew the asker was associated with Northwest Lagotto. Every money query was therefore re-run in a memory-off temporary session. Both conditions are reported below, separately, and where they disagree the disagreement is the finding.
Map cards are not the datum
Entity cards and map results may lean on geolocation, which a reader in a different place cannot reproduce. The text of each answer is the datum throughout. Where a card is mentioned, it is context, not evidence.
Two captures carry superseded language
Two of the eleven captures still quote health-testing language the site had already corrected before publication. They are reported as captured rather than re-run, because the lag between a corrected entity and a stale external record is itself one of the findings. See the correction log below.
Web search and AI engines are different instruments
Ten standard web searches ran the same day. They are a proxy for framework fit, not an AI-engine result, and they are reported in their own section rather than blended with the eleven captures. Only the eleven captures carry engine, account condition and timestamp; a standard web search has no account condition to log.
AI-engine captures
Standard web searches
Ten searches, July 11, 2026: standard web search, not AI-engine queries. Reported separately because they measure something different.
“Lagotto Romagnolo breeder Pacific Northwest”
NWL appears twice on the first page — homepage and About page — alongside five genuine regional competitors. The synthesized answer names NWL first and with the most specific detail: Puppy Culture, CHIC testing, the ten acres, the years with the breed. A real field, not an empty one.
“Best Lagotto Romagnolo breeder Washington state”
The same pattern holds: NWL first, ahead of four named alternatives.
“Northwest Lagotto reviews”: the unflattering one
The synthesized answer mostly pulls testimonial language from NWL's own site, plus a Yelp listing and an unaccredited BBB profile. Most of what is currently findable about NWL's reputation traces back to NWL's own domain, not to an independent third-party review platform. That is a real gap, not a rounding error, and it isn't one a schema fix solves by itself.
A related, smaller finding: the same search surfaces lagottonw.com, a different, unrelated breeder in Vancouver, Washington, confirmed as a distinct business, not a redirect or an NWL property. It is a name-confusion risk for a buyer typing carelessly, not a technical problem, and it is the kind of gap an entity graph doesn't close on its own.
“The only breed developed for truffle hunting”: the negative finding
NWL isn't cited at all. The synthesized answer draws on Wikipedia, the LRCA, and general breed-authority sources. This is the finding that sharpens the claim rather than undermining it: NWL wins geography-plus-breeder queries, the ones an actual buyer asks. It does not win generic breed-education queries, the ones with no purchase intent behind them.
The seven-layer audit
Clarion's 130-firm research report specifies seven layers a complete entity graph needs. Zero of the 130 audited RIA firms have all seven. The comparison table is on the case study itself; what follows is the detail behind the two rows that need more than a checkmark, plus the wider-evidence caveat.
Regulatory verification: the most striking row
Clarion's research calls this the most striking finding in the study, precisely because it is the cheapest layer to build and the one almost nobody builds: zero of 130 audited RIA firms link a single regulatory body from their own site. Northwest Lagotto links three — AKC Marketplace, LRCA membership, and per-dog OFA records — not because dog breeding carries more regulatory infrastructure than registered investment advice (it doesn't; there is no SEC-equivalent for Lagotto breeders), but because the equivalent authorities exist and somebody bothered to link them.
One qualification a reader should apply themselves: the BBB profile in NWL's sameAs is unaccredited, and the case study's own reputation finding says so. It is a link to a real external record, not a credential.
Propagation: where the audit process is visible
Clarion's July 2026 audit of NWL's schema — the same audit standard applied to client sites, run on the studio's own reference asset — found the canonical Organization, Person and Website entities defined inline, redundantly, on multiple pages instead of declared once and referenced by @id everywhere else. The kind of drift that accumulates quietly on any site that grows. It was remediated across the full site the same month, before the case study published, and the check that caught it now runs as a standing part of the audit protocol. Entity graphs aren't built once. They're maintained, and the maintenance is checkable too.
The wider category evidence, and how much weight it bears
Industry analyses — not peer-reviewed, and held to a lower confidence than Clarion's own audit — suggest that roughly sixty-one percent of ChatGPT-cited pages carry rich schema markup, against roughly twenty-five percent of standard search-result pages, and separately suggest a thirty-six percent lift in AI-summary appearance tied to schema deployed sitewide rather than on a single page. Read those two numbers as directional, sourced to technical-SEO agency content rather than independent research, and not as findings Clarion verified itself. The 130-firm audit is the number worth standing behind; the industry figures merely point the same way.
The correction log
A case study that only shows what works isn't evidence. Here is what the audit found when it was pointed at the studio's own reference asset, and what happened next.
Two live claims that were wrong
NWL's internal audit corrected a stale summer-litter announcement that contradicted an already-closed litter elsewhere on the site, and language describing CHIC certification as complete when the annual eye exams a current CHIC number requires were still being scheduled. Both corrections are deployed to production. I checked by fetching the live site directly on July 11, 2026, rather than trusting my own memory of having fixed them.
The homepage now reads “Autumn 2026 Litter, Inquiries Open.” The health-testing language now says exactly what is true: DNA-clear or DNA-carrier status confirmed for the two breed-specific conditions, OFA hips completed, eye exams pending, certification expected this summer.
The lag, watched in a single day
Search engines, that same morning, were still serving the pre-correction claims — the old litter status, the “CHIC-certified” language the corrected page no longer uses. Two of the AI-engine captures above still carry that stale phrasing. But by the end of the same day's capture session, Google's organic snippet for NWL already read from the corrected page: “health tested, publicly verifiable at ofa.org.”
The entity changed, the external record trailed it by weeks, and then it caught up — morning and evening of the same day, both states dated and screenshotted. That is what the lag looks like when you can actually watch it move. The owner of an entity graph doesn't control that lag. What the owner controls is which version of the record the recrawl finds.
A finding that cuts against Clarion, not NWL
NWL's robots.txt carries an AI-crawler allow-list naming GPTBot, ClaudeBot, PerplexityBot, Google-Extended and more, explicitly welcomed rather than blocked by default. Clarion's own site had the opposite problem until that same week, when a default-on Cloudflare setting quietly blocking those same crawlers was found and fixed during the audit that produced this case study. The subject of this piece had the plumbing right before the studio auditing it did.
Verify it yourself
Everything above is checkable without taking my word for any of it.
Read the schema. View source on northwestlagotto.com and read the JSON-LD in the head directly. Compare the entity declarations to the seven-layer table on the case study.
Re-run the queries. Every query above is reproduced verbatim so it can be run again. Run them on whatever date you are reading this and compare to what is reported here as of July 11, 2026. Results drift, and an evidence file worth trusting should tell you how to catch it doing so.
Check the external records. The AKC Marketplace listing, the LRCA membership, and the per-dog OFA records at ofa.org can each be checked against the site's own claims.
And read this with the interest disclosed. I own Northwest Lagotto and I built the site. Clarion Studio sells the work this file documents. Nothing here is independent evidence. What it is instead is checkable, which is a different and lower claim, and the only one this file makes.