Photographs do not get cited by AI search. The paragraphs next to the photographs do. That is why FAQ schema matters more than almost any other on-page change you can make for AI visibility, and why a working corporate photographer is the one writing this post. FAQ schema is the single highest-impact change you can make on your service pages for AI search visibility, and it takes about 20 minutes per page to implement. I am going to give you the exact JSON-LD I ship on every page on this site, the question-writing rules I follow, and the answer-length sweet spot that AI systems consistently prefer. Then I will show you how to verify it works in five minutes using a free Google tool. This post is a spoke under my AI search visibility cluster.
Why FAQ schema beats every other structured data type
More than a dozen schema markup types are relevant for a service business: Organization, LocalBusiness, Service, Product, Article, Review, BreadcrumbList, ImageObject, VideoObject, and a handful of vertical-specific types. They all matter to varying degrees. FAQPage is the one with the strongest signal for AI citations.
Per multiple 2026 analyses of AIO citation patterns, content with proper FAQPage schema sees roughly a 2.5x higher chance of being cited in AI search answers compared to the same content without it. GPT-4's accuracy interpreting page content jumps from 16% to 54% when structured data is present. Pages with three to four layered schema types (LocalBusiness plus FAQPage plus Service plus BreadcrumbList, for example) get cited at roughly 2x the rate of pages with one schema type.
The reason FAQ schema beats the rest is that it packages information in the exact format AI systems prefer. A question paired with a 134 to 167-word answer is the unit retrieval models are trained on. Wikipedia, Stack Overflow, and Quora all structure content this way, and those domains dominate AI citations across every platform. Your service page in question-and-answer format is the closest a small business can get to looking like the sources AI was trained to trust.
The five rules for questions AI will actually cite
Not every question gets cited. After looking at the question patterns inside the AI answers I have logged from my own searches, five rules separate questions that get pulled into AI answers from questions that get ignored.
Use the actual phrasing your customer would type or say. Not the phrasing you wish they used. "How much do corporate headshots cost in St. Louis?" beats "What is the investment for executive portrait sessions?" The first is what someone types into ChatGPT. The second is brochure language. AI systems pattern-match against real-world question phrasing. If your question doesn't sound like something a customer would actually ask, it won't match.
Lead with the W: who, what, when, where, why, how, how much. Questions with a clear interrogative at the front get parsed cleanly by retrieval models. Questions phrased as statements ("Pricing for our team headshots") confuse the parser and rarely get cited.
Be specific to your service area. "How much does a corporate headshot photographer cost?" is generic and competes with every photographer on the internet. "How much does a corporate headshot photographer cost in St. Louis?" anchors to a service area, narrows the relevant comparison set, and increases the odds of being cited when someone searches with location intent.
Pull the questions from your inbox, not your imagination. Open the last 50 emails from prospects. Read the questions verbatim. Those are your FAQ items. The exact wording, including any awkward phrasing, is more valuable than your polished version. Real questions match real searches.
One question per FAQ item. Don't combine "How much do headshots cost and how long does it take?" into a single question. Split it. AI systems treat each Question/Answer pair as an independent retrieval unit. Two cleanly-scoped questions cite twice as often as one bundled question.
The 134-to-167-word answer length and why it works
Research published in 2026 on AI Overview citation patterns identified a clear sweet spot for answer-block length: roughly 134 to 167 words per self-contained answer. Shorter answers don't carry enough information density to be useful. Longer answers get truncated or split, and the AI loses the thread.
A 134-to-167-word answer is long enough to include a specific number, a location, a credential, and a verifiable claim. Short enough to lift cleanly into an AI answer with attribution. This is not a theoretical preference. It is the format the retrieval models were trained to prefer, because it matches the average paragraph length on the high-trust sources they learned from.
When I rewrite a page's FAQ section, I count words. Every answer lands between 134 and 167. If I am over, I cut. If I am under, I add a specific detail (a real client industry, a price range, a typical project timeline). The discipline matters because the difference between a 90-word answer and a 150-word answer is the difference between not being cited and being cited.
My exact FAQPage JSON-LD template
Here is the exact JSON-LD format I ship on this site. The component is reusable, but the underlying schema is plain JSON-LD that works in any framework, any CMS, or hand-coded HTML.
```json
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "How much do corporate headshots cost in St. Louis?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Corporate headshot pricing in St. Louis ranges from $250 for a single individual session to $35-$95 per person for team programs of 10 or more, depending on group size, location, and deliverables. My on-location team rate at 25-50 people is $65 per person, all-inclusive: tethered shooting, expression coaching, retouching, and final files formatted for LinkedIn, your team page, and print. Sessions take roughly eight minutes per person on a properly-scheduled day, and most teams have proofing galleries within one business day of the shoot. Travel is included for organizations within an hour of St. Louis. For multi-office programs, I quote a flat travel line item that distributes across all the headshots, which usually keeps the per-person rate competitive with what you'd pay a local photographer in each city."
}
},
{
"@type": "Question",
"name": "How long does a corporate headshot session take?",
"acceptedAnswer": {
"@type": "Answer",
"text": "On a properly-scheduled team headshot day, each person spends about eight minutes in front of the camera, start to finish. That includes positioning, expression coaching, capturing two to three distinct looks, and reviewing every frame on the tethered 32-inch monitor before they leave. The full block I reserve per person is 10 minutes to give buffer for transitions and walk-ins. A team of 40 fits comfortably into a single day with breaks, and a team of 80 fits into one extended day or two half-days. Setup before the first appointment takes 45 minutes. Teardown takes 20 minutes. The actual photography is fast because the system is built for it, not because I am rushing anyone."
}
}
]
}
```
This goes inside a `<script type="application/ld+json">` tag in the `<head>` of the page, or anywhere in the body if your CMS doesn't allow head-tag injection. Modern Next.js sites can use a JSON-LD component (mine is at `src/components/seo/JsonLd.tsx`) that renders the script tag. WordPress sites can use Yoast SEO, Rank Math, or a custom function in `functions.php`. Static sites just paste the script tag into the HTML.
The critical detail: the questions and answers in the JSON-LD must match what appears visibly on the page. Google has explicitly stated that hidden FAQ content (in the schema but not visible on the page) violates structured data guidelines. Render the same questions and answers in human-readable HTML on the page itself. The schema is the AI-readable version of the same content.
Want the full AI-Visual Branding Package?
FAQ schema is one piece. The shoot day, the multi-modal pages, the AI search setup, and the quarterly citation tracking are the rest. See the full engagement.
See the packageHow to verify it works in five minutes
Google's Rich Results Tool is the canonical verification path. Free, instant, definitive.
Go to search.google.com/test/rich-results. Paste your page URL. Click Test URL. Within 30 seconds, you will see a results screen showing detected structured data types. "FAQPage" should appear in the list with a green checkmark. Click it to see each Question/Answer pair the tool detected and verify they match what you intended.
If you see warnings (yellow triangles), read them carefully. Common issues: question text appearing in the schema but not on the page, answer text exceeding the character limit, missing required fields. Fix the warnings before moving on.
If you see errors (red X), the schema won't be eligible for rich results at all. Errors usually mean malformed JSON, missing closing brackets, or required schema properties left out. Validate your JSON-LD at validator.schema.org for a more granular error report.
Once the Rich Results Tool shows green checks, the FAQ schema is live and AI search systems will pick it up the next time their crawlers retrieve your page. For ChatGPT and Perplexity, that can be within hours. For Google AI Overviews, it follows the same indexing schedule as the rest of your site.
What I ship on every service page on this site
Every service page on henrydavidphotography.com ships with FAQPage JSON-LD generated by a single shared React component. The questions are pulled from real client emails and sales calls. The answers are 134 to 167 words. The questions are visible on the page itself in expandable accordions, and the JSON-LD mirrors the visible content exactly.
The service pages currently running this include corporate photography, professional headshots, and every industry vertical (healthcare, financial services, legal, technology, construction, agriculture, professional services). The conference headshot booth page, the LinkedIn headshots page, the personal branding page, and the modeling portfolios page each have their own FAQ blocks tuned to the questions specific to that audience.
The content of the questions is what matters most. Generic photographer-industry questions get less traction than questions tied to a real customer pain point. "What if our new hires need to match the existing team's headshots in 2027?" gets cited more than "Do you offer team photography?" because the first one names a specific scenario that maps to a customer's real anxiety, and the answer can be a concrete process description with timing.
Common mistakes that kill the citation
After shipping FAQ schema across every service page on this site and watching what gets cited and what doesn't, five mistakes show up over and over on other small-business sites I audit.
Stuffing keywords into questions. "What is the best corporate headshot photographer in St. Louis Missouri for executives, professionals, and teams?" reads like an SEO 2014 attempt at keyword density. Real customers don't type that. Retrieval models recognize the pattern and downrank it. Per a 2026 study cited in GEO research, keyword-stuffed content underperformed baseline content by about 10% in retrieval tests. Write like a human. Pick the one query phrasing your customer actually uses.
Answers that don't actually answer. "It depends on your needs. Contact us for a custom quote." is not an answer. It is a deflection. AI systems will skip it because there is no verifiable, citable information to lift. Even if your real answer is genuinely "it depends," give the depending factors and the typical range. "Pricing depends on team size and location. Most multi-office programs land between $50 and $95 per person all-inclusive." That gets cited. "Contact us for a custom quote" doesn't.
No specific numbers. Citations love numbers. Prices, durations, team sizes, response times, deliverable counts. The more specific numerals in your answer, the more likely it is to be lifted into an AI response. "We work with teams of all sizes" is filler. "We have photographed teams ranging from five-person partnerships to 800-person enterprise rollouts across multiple offices" is citable.
Hidden FAQ schema (in the JSON-LD but not on the page). Google's structured data guidelines explicitly state that schema content must match visible page content. Schema-only FAQs are a manual-action risk and a guaranteed credibility loss. Always render the FAQ visibly on the page in addition to the JSON-LD.
No update cadence. Pricing changes. Service offerings shift. Locations expand. If your FAQ schema is two years old and contradicts your current website copy, AI systems pick up the contradiction and trust your content less. Review every FAQ block quarterly. Refresh the answers with current numbers.
What to do today
If you've been considering FAQ schema and haven't pulled the trigger, do this in the next 30 minutes.
Pick your highest-revenue page. The one that drives the most contact form submissions or phone calls. Open your inbox and pull five questions that prospects actually asked you in the last 60 days. Write 134 to 167-word answers to each. Wrap them in the JSON-LD template above. Inject the script tag into the head of the page. Test in Google's Rich Results Tool. Done.
The entire workflow takes about 30 minutes if you're using the questions and answers you already have in your email. Tomorrow, do it on your second-highest-revenue page. By the end of the week, every customer-facing service page on your site has FAQPage schema and you've moved your AI citation odds up by roughly 2.5x.
If you want me to do this on your site rather than DIY, get in touch and we'll talk it through.