For a wedding photographer, a portrait studio, a commercial product photographer, or a fine-art print seller, the website's images are the literal product. The schema markup that describes those images carries real weight in three different audits — rights enforcement, AI image-search citation, and AI training opt-out signaling. And until round 8, the Mega Analyzer had a quiet gap: it only ran the image-rights schema checks when an ImageObject block was already present in the page.
That meant a photographer site with zero image schema sailed through the structured-data bucket untouched. The cleanest failure mode in audit tooling is a green check on a check that wasn't even run.
The three audits image schema serves
The first audit is rights enforcement. When a wedding magazine, a vendor blog, or an AI-generator dataset uses a photographer's image without permission, the disputes that go well are the ones where the original site clearly declared ownership and license terms at the markup level. Schema.org has explicit fields for this: copyrightHolder (a Person or Organization), copyrightNotice (the human-readable © string), creditText (how the photo should be credited), license (a URL to the license terms page), and acquireLicensePage (a URL where rights can be requested or purchased).
Google's image-license metadata documentation describes these fields and how Google Images surfaces them in the lightbox. Bing and DuckDuckGo Images use similar parsing. Pinterest reads them. AI training scrapers either honor them or face a clearer chain of evidence in subsequent disputes — both outcomes are better than the alternative.
The second audit is AI image-search citation. When ChatGPT or Perplexity is asked "show me wedding photographers in Aspen who shoot 35mm film," the answer-engine can either name specific photographers with linked galleries (citation-worthy) or describe the genre generically (not useful for the photographers competing in that space). The citation outcome depends on whether there's a citable entity to point at. A Photograph JSON-LD block with contentLocation, creator, datePublished, and name is exactly that citable entity.
The third audit is AI training opt-out signaling. The Spawning ai.txt spec, Cloudflare's Content-Signal directive, and several emerging IETF drafts all give photographers a way to declare their training preferences. Schema-level license fields compose with those signals: when a photographer has license: https://creativecommons.org/licenses/by-nc-nd/4.0/ on the image schema and Content-Signal: ai-train=no on the robots.txt, the combined signal is clearer than either alone. Compliant scrapers honor it. Non-compliant ones leave a clearer evidence trail.
Why the existing check didn't catch this
The Mega Analyzer's imageMetaGSC family of checks (which validates copyrightNotice, license, acquireLicensePage, and creditText against any ImageObject block found on the page) was written for the case where ImageObject schema already exists and just needs better field coverage. That's the right shape of check for an editorial publication or a stock-photo site where ImageObject is a known requirement.
For a photographer's marketing site, though, the more common failure is the absence of any image schema at all. The site has a portfolio, a Showit/Squarespace/Pixieset template, beautiful galleries, and zero JSON-LD describing any of it. The existing check passes by definition because there's nothing to validate. The gap is structural.
How round 8 fixes it
The new check uses two converging signals to identify a photographer site:
The first is platform fingerprint. If the vendor platform is one of Pixieset, SmugMug, Format, Pic-Time, ShootProof, or Zenfolio — all of which are photographer-focused SaaS — the site is treated as a photographer site by default. (This composes nicely with the robots.txt platform fallback that ships in the same round.)
The second is copy fingerprint. The page title, H1s, and meta description are scanned for the photographer-vocabulary regex, which catches all the common subgenres in one pass: wedding, portrait, family, newborn, maternity, engagement, elopement, boudoir, headshot, commercial, product, fine-art, documentary, editorial, brand, food, landscape, fashion, architectural, real-estate, equestrian, pet, senior, graduation. The regex is broad on purpose because the cost of a false-positive is low (one extra schema recommendation) while the cost of a false-negative is a photographer site sailing past with no image-rights check at all.
When either signal fires AND no ImageObject AND no Photograph schema is detected, the analyzer emits a warning with a concrete fix recipe: add ImageObject schema with copyrightHolder, copyrightNotice, creditText, license, and acquireLicensePage on at least the top 5–10 portfolio pieces. The recipe is opinionated — five to ten covers the citation-worthy work without forcing the photographer to mark up every single gallery image, which would be infeasible on a 200-image portfolio.
The general pattern
This is the third round-8 addition that shares the same shape: gate a check by vertical, not just by what's already present in the markup. The first addition gated the AGENTS.md and security.txt checks by platform (you can't publish those on Pixieset; don't false-warn). The second gated the bot-challenge severity by platform (Pixieset can't toggle Cloudflare directly; route remediation to vendor support). The third gates the image-rights check by vertical (photographer sites need this; brochure HVAC sites don't).
The lesson generalizes: every audit tool eventually trips on its own one-size-fits-all assumption. The path out is to compose the existing checks with cheap classifiers — platform, vertical, copy signals, response headers — that scope which findings should fire as warnings versus info versus suppressed entirely. False-positive fatigue is the single biggest reason teams stop trusting an audit tool. Vertical-aware gating is one of the cheapest ways to fight it.
Related reading
- Robots.txt as a platform fingerprint — the platform-detection underpinning that lets this check route differently for Pixieset vs Showit vs a self-hosted site.
- Vercel BotID and Cloudflare Bot Fight Mode silently block AI search — the second round-8 addition; photographer sites are over-represented in the affected population because Pixieset's default Cloudflare configuration challenges most AI crawlers.
- One URL, every audit in one pass — the Mega Analyzer — the umbrella post.
For photographers building or rebuilding their site, The $20 Dollar Agency has a chapter on the under-$100 AI stack for service businesses where the photographer's own portfolio is the SKU. Chapter 8 walks through schema and image-rights wiring specifically for visual-product service businesses.
