The highest-ROI single action
The structured pages AI engines quote word for word
FAQ schema hands an engine a question and its answer, already paired and labelled. It is the cheapest item on the fix list, and the one most sites skip.
Every plan queries all 5. See how each engine picks who it names
The number behind this page
sites we crawled during a scan carry no FAQ schema at all, and those 39 are not a random sample: every one of those firms ran a scan because someone suspected it was missing from AI answers.
The gap this closes
- An engine reading marketing prose has to infer what your answer would be.
- 26 of 39 readable crawls we ran carry no FAQ schema at all, and those firms all suspect they are absent, so this is not a random sample.
- That is two thirds of a group already worried about being invisible.
- It is the cheapest item on the fix list and the one most often skipped.
What you end up with
- Finished pages with the markup already inside them, not a checklist telling you to add it.
- Question and answer pre-paired and labelled, so nothing is left to inference.
- Valid FAQPage JSON-LD that matches the visible text exactly.
- A fix measured in an afternoon, not a rebuild.
How it works
Three steps.
The clusters come from losing queries
We do not brainstorm queries. The scan already recorded which buying queries an engine answered by naming a competitor. The clusters are drawn from those, in buying-intent order, so the first page you publish is aimed at a question that decides a shortlist.
Each answer has to stand on its own
An engine quotes one answer, not your page. So every answer is written to survive being lifted out of context: specific, self-contained, and phrased the way a buyer phrases it. Compliance detail, company size and tech stack go in, because the qualifying question turns on them.
The markup matches the page, and the next scan referees it
Schema that claims content a reader cannot see gets treated as spam. So every question and answer in the JSON-LD renders on the page as well. The Content Engine publishes the page to your CMS. The next scan re-crawls your site, reports whether the markup is live, and asks the original query again on all five engines.
What you get on each plan
You get finished pages with the markup already inside them, not a checklist telling you to add markup. This is what lands, per plan, with the monthly volume read live from the plan configuration.
Track
Free- No drafting on this tier
- The query clusters where somebody else was named
- Ranked by buying intent
- The crawl reports whether you carry FAQ markup today
- Next scan tells you if the markup went live
Fix
$99/mo- FAQ pages built monthly against your top clusters
- FAQ blocks ride along inside every other draft
- Valid FAQPage JSON-LD matching the visible text
- Each page shows the answer that lost the query
- Publishes through your CMS, nothing without your say-so
Dominate
$499/mo- Enough volume to blanket a category, not pick at it
- Clusters chosen across the wider query budget
- Video and newsletter reuse the same answers
- One cluster answered on site, YouTube and newsletter
- Contested clusters get built before empty ones
Rule
$1999/mo- Dominate's numbers, with our team doing the work
- We check the markup validates before it ships
- Publishing to your site only if you grant it
Rule is portfolio level and sales-assisted, and it is the right conversation if you are running this across more than one brand.
The proof, including the parts that flatter nobody
Our own scans are why this sits at the top of the fix list. Across 46 completed reports, on firms that arrived already suspecting they were missing, the best any engine managed was naming the brand in 7.4% of the buying queries we asked it.
The others sat between 0.4% and 4.0%, on the same firms that suspected they were missing.
Being listed somewhere is not the same as being retrievable, and we have the harder version of that number. We asked four engines to recommend managed IT providers across 15 North American regions.
In that managed IT study, 139 of the 228 firms listed in those regions' own Clutch directories were named by no engine at all.
Structure is what the engines can read. A directory entry is what they can ignore.
Read the MSP AI Visibility Report 2026 →A worked example: the block and the markup it ships with
One pair from a page built for a firm losing the query asking whether a small clinic can pass a HIPAA audit with an outsourced IT provider. The firm is illustrative; the structure is what every page ships with. The answer is written to survive being lifted out of the page, because that is what happens to it.
What the reader sees
- Q. Can a 20-person clinic pass a HIPAA audit with an outsourced IT provider?
- A. Yes, and the audit turns on evidence rather than headcount. A provider has to supply a signed business associate agreement, an asset inventory, access logs going back six years, documented annual risk analysis, and encryption at rest and in transit. Ask for those five artefacts before you sign. A provider who cannot produce them for an existing client will not produce them for you.
What ships in the page head, matching word for word
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [{
"@type": "Question",
"name": "Can a 20-person clinic pass a HIPAA audit with an outsourced IT provider?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Yes, and the audit turns on evidence rather than headcount. A provider has to supply a signed business associate agreement, an asset inventory, access logs going back six years, documented annual risk analysis, and encryption at rest and in transit. Ask for those five artefacts before you sign. A provider who cannot produce them for an existing client will not produce them for you."
}
}]
}Three properties make that answer quotable. It commits in the first clause, so a system extracting one sentence still gets your position. It carries five checkable specifics, which separate an answer worth citing from one worth skipping. And it names the qualifying detail a buyer holds in their head, the clinic size and the audit, so it matches the question a real person typed.
The page it lives on carries the rest of the cluster: the business associate agreement, the six-year log retention, what happens during a breach investigation. One page, one cluster, every pair marked up, every pair visible.
How it actually works
The mechanism, for anyone who wants it. Open a card to read the detail.
What the markup does to an answer
A page of marketing prose forces a reading system to infer four things. Every one can go wrong, and a system with a cheaper option takes it:
Read the detail
- Which sentence was meant as an answer.
- Which question that answer belongs to.
- Where the answer stops.
- Whether the surrounding paragraph changes its meaning.
FAQPage markup removes all four decisions:
- Each question is a labelled string, and so is each answer.
- The pairing is declared instead of inferred.
- A machine assembling a response gets a quotable unit with a source attached and no ambiguity about where it ends.
That is the whole trick, and it is why this sits at the top of a fix list rather than in the middle of one.
The second effect is quieter. Structured data also tells a reader what kind of thing your page is about. An Organization block, a Service block and a set of question and answer pairs together describe a firm, what it sells, and what it says about the questions buyers ask. Systems that decide who to name are looking for exactly that description.
Google cutting FAQ rich results made this more valuable
In 2023 Google restricted FAQ rich results to a short list of government and health sites. A lot of teams read that as the end of the format and pulled the markup out.
Read the detail
The rich result went away. The parsing did not. The same markup now feeds AI Overviews and AI Mode, the surface that answers the buying query a blue link used to answer.
So the market position is unusual, and worth being blunt about:
- A format got written off for the wrong reason.
- A large share of sites removed it or never added it.
- The systems that matter most for vendor selection still read it.
- Cheap, unclaimed and machine-readable is a rare combination.
The four ways markup gets ignored
None of the four is hard to avoid. All four are common, and all four turn work you already paid for into work that does nothing.
Read the detail
- Schema that overstates the page. Markup claiming an answer a reader cannot see is treated as spam, and the penalty lands on all of your structured data, not on the one block that lied. Everything we ship renders the same text it declares. Our own test suite compares the JSON-LD against the visible page byte for byte before a deploy.
- Invalid types. An invented type name reads as nothing at all. Our own report validation had to start stripping schema recommendations that named types Schema.org does not define.
- Placement. Markup written into the page by a script the crawler never runs is markup nobody sees, so it belongs in the served HTML.
- Nesting. Blocks buried inside a graph get missed by naive parsers. We found that defect in our own crawler, when a valid Offer block on our pricing page read as absent.
See where you stand first
Run a free scan and find out which of these gaps you actually have.
Get your free auditFrequently asked questions
What is FAQ schema?
A structured-data format defined by Schema.org that labels question and answer pairs so a machine reads them without guessing. Search engines and AI engines both parse it, and the AI benefit is now the bigger of the two by some distance.
Do the answers have to match what is visible on the page?
Yes, always. Markup that claims content a reader cannot see is treated as spam and can get every bit of your structured data ignored. Every page we ship renders the same questions and answers that its schema declares.
Can I add FAQ schema myself with a plugin?
The markup, yes: plugins emit valid JSON-LD and that part is solved. The part that moves visibility is knowing which questions to answer, phrased the way buyers ask an engine, with answers specific enough to quote. That comes from scan data, not from a plugin.
Does it still help with Google?
Yes, for a different reason than in 2022. Google cut most FAQ rich results in classic search, and the same markup still feeds AI Overviews and AI Mode, which is where the value now sits.
How many FAQ pages does a site need?
Enough to cover the query clusters your buyers ask, which for most B2B firms is a dozen or more. We build them in buying-intent order, so the pages that can win a shortlist go before the pages that explain a category.
How do I know it worked?
The next scan re-crawls your site and reports whether the markup is live, then asks the query that prompted the page again on every engine. A fix you cannot verify is a fix you are taking on faith.
Sources and further reading
- SE Ranking, structured data in AI answers (via Search Engine Land): 71% of pages ChatGPT cites include structured data; 65% for Google AI Mode.
- Schema.org FAQPage specification: The machine-readable format AI engines parse to extract question-answer pairs.
Related solutions