Key facts
- Six parts, each checkable. Every letter comes with concrete checks and a tool that runs them.
- Works across assistants. ChatGPT, Claude, Gemini, Perplexity, and Google's AI Overviews and AI Mode.
- No special markup required. Google says there are no additional requirements, special schema or "AI text files" needed to appear in AI Overviews or AI Mode. Google Search Central
- JavaScript is the most common blind spot. An analysis of AI crawler traffic found ChatGPT's and Claude's crawlers fetch JavaScript files but don't run them, so content that only appears after scripts run is invisible to them. Vercel
- One answer, many searches. Google's AI features may issue "multiple related searches across subtopics" to build a single response. Google Search Central
How an AI answer gets built, and where pages drop out
When a buyer asks an assistant for a recommendation, the assistant usually doesn't rely on memory alone. It searches, often several times with different wordings, picks a handful of pages, pulls passages out of them and writes an answer, sometimes with links to its sources.
A page can drop out at three points. It can fail to be found, because the crawler is blocked or can't read the content. It can fail to be chosen, because it doesn't answer the specific sub-question or nothing independent backs it up. Or it can be chosen but misread, because its facts are vague, outdated or contradicted elsewhere. QUOTED covers all three.
Can the crawler fetch and read the page?
EDoes it answer the sub-question, and do others vouch for it?
Q · OAre the facts clear, specific, current and consistent?
U · T · DQuestion-first
Put the answer in the first two sentences, then answer the questions a buyer asks next.
Why it matters
Assistants rarely search for your exact keyword. A prompt like "best project planning tool for a 20-person agency" gets broken into several searches: pricing, alternatives, integrations, reviews, limits. A page that answers the main question straight away and covers those follow-ups gives the assistant fewer reasons to go elsewhere, and more passages it can use.
How to apply it
- Open with a 40–60-word direct answer: what it is, who it's for and when to choose it.
- Make each H2 a question buyers actually ask. Take them from sales calls, support tickets, Search Console queries and autocomplete, not from a keyword tool alone.
- Cover the usual follow-ups on the page or one click away: price, alternatives, integrations, limits and set-up time.
- Publish a comparison page for each main competitor and an "alternatives" page for your category. They're what an assistant needs when someone asks it for alternatives.
Common mistakeThree paragraphs of scene-setting before the answer. The assistant will often take the answer from a page that gets to the point.
Unmistakable entity
Make it impossible to confuse who you are, what you do and how you relate to everything around you.
Why it matters
Models connect facts through entities: your company, your product, its category, your competitors, the tools you integrate with. If your name is also an everyday word or another company's name, facts get mixed up. If you never say outright what you replace or connect to, the model has to guess, and often guesses wrong.
How to apply it
- Name your category in plain words on the homepage and product pages: "Acme is project planning software for agencies."
- Write the relationships as sentences: "An alternative to X", "Integrates with Y", "Built for Z".
- Add Organization, Product or SoftwareApplication, and FAQ structured data, and link your official profiles with
sameAs. - Mark up only what's visible on the page. Google's own guidance is to make sure "structured data matches the visible text on the page".
- If your name is ambiguous, say what you are every time it matters, including in page titles.
Common mistakeSchema that claims more than the page says, such as ratings or prices that aren't shown. It doesn't help, and it can count against you.
Other people's proof
What other people say about you counts for more than what you say about yourself.
Why it matters
When an assistant builds a shortlist, it leans on sources that look independent: review sites, directories, comparison articles, press coverage and forum threads. If you're missing from them, you're less likely to be named. If they keep repeating an old complaint, the assistant will repeat it too.
How to apply it
- Claim and complete your profiles on the review sites for your category (for software, sites like G2, Capterra and TrustRadius) and link to them from your site.
- Ask customers for reviews steadily, rather than in one burst a year.
- Find the comparison and "best of" articles that rank for your category, and work to be included.
- Take part honestly in the communities where your buyers ask questions.
- Answer recurring complaints on your own site, with a clear page on pricing changes, support or the roadmap, so there's something better to quote.
Common mistakeOnly working on your own website. For shortlist questions, third-party pages often matter more.
Truth you can check
Specific, checkable facts get quoted. Vague claims get paraphrased, often wrongly.
Why it matters
When an assistant writes an answer, it prefers statements it can lift and attribute: a price, a plan name, a limit, a date, a number with a source. "Affordable plans for every team" gives it nothing to quote, so it fills the gap from somewhere else, such as an old review, a competitor's comparison page or its training data.
How to apply it
- State prices, plan names, limits and key numbers in plain text, with the currency and billing period.
- Back numbers with a source, and link to it.
- Give each section one or two sentences that make sense when quoted on their own.
- Ask the assistants directly. Check what ChatGPT, Claude, Gemini and Perplexity say about your pricing, integrations and security, and compare it with your own pages.
Example
Before: "Flexible pricing that grows with your team."
After: "Acme Pro costs $39 per user per month, billed annually, or $45 billed monthly. Every plan includes unlimited projects."
Extractable
If a crawler can't read it, or a model can't lift it cleanly, it won't be quoted.
Why it matters
There are two separate problems here. The first is access. AI crawlers have to be allowed in your robots.txt, and they have to find the content in the HTML. ChatGPT's and Claude's crawlers don't run JavaScript, so prices, specs or text that only load by script are invisible to them. Gemini is the exception, because it uses Google's rendering. The second is shape. Answers are put together from passages, and a passage that only makes sense with the paragraph before it gets skipped or misquoted.
How to apply it
- Check your robots.txt crawler by crawler. Decide separately about AI search crawlers (which can send you visitors) and training crawlers.
- Make sure the content that matters (prices, plans, specs, docs) is in the server-rendered HTML.
- Write in self-contained blocks: a clear heading followed by a few short paragraphs that stand on their own.
- Start long pages with a short summary, and use tables and lists for anything people compare.
- An llms.txt file is optional. It's a proposed convention, and Google says it doesn't need one, but it costs little to add.
Common mistakeA pricing page that looks perfect in the browser but serves an empty table to crawlers, because the numbers load from a script.
Dated & consistent
Show when a page was last updated, and say the same thing everywhere.
Why it matters
Assistants favour current information and get confused by contradictions. If your pricing page says €49, a review profile says €39 and a partner page says "from €29", the assistant may hedge, average them or pick the wrong one. The more places your facts appear, the more places they can drift.
How to apply it
- Show a visible "Last updated" date on pages where freshness matters, and change it only when the content changes.
- Keep one fact sheet: prices, plans, limits, integrations, certifications, locations and founding date.
- Fix the copies that drift: review profiles, directories, partner listings, old blog posts and docs.
- Keep the
lastmoddates in your sitemap accurate. - Log the changes you make, and check afterwards whether they moved anything.
Common mistakeUpdating the pricing page and forgetting the five other places that quote the old price.
The QUOTED checklist
Run this against any page that matters to buyers. Start with pricing, product, comparison and alternatives pages, integrations, security and docs.
| Letter | Check | Passes when |
|---|---|---|
| Q | Direct answer up top | The first 60 words say what it is, who it's for and when to choose it |
| Q | Follow-up questions | Price, alternatives, integrations and limits are answered or linked |
| U | Category and relationships | Category, alternatives and integrations are stated in plain sentences |
| U | Structured data | Organization / Product / FAQ schema present and matching the visible text |
| O | Review sites | Complete profiles on your category's main review sites, linked from your site |
| O | Recurring complaints | Your own pages answer the complaints AI repeats |
| T | Specific facts | Prices, plans and limits stated in plain text, with currency and period |
| T | AI accuracy | The main assistants give your price and features correctly |
| E | Crawler access | AI search crawlers allowed in robots.txt |
| E | No JavaScript dependency | Key content present in the HTML without running scripts |
| D | Visible date | "Last updated" shown, and accurate |
| D | Consistency | The same facts on your site, docs, review profiles and partner pages |
Audit a page with AI
Paste this into ChatGPT, Claude, Gemini or Perplexity with a page's URL or text. It's a quick first pass, not a substitute for checking what crawlers actually receive. A model reading a URL may run scripts that the crawlers don't.
Review this web page for how well AI assistants can find, trust and quote it, using the QUOTED framework. For each letter, give a score from 0 to 2, quote the evidence from the page, and name the single most useful fix. Q, Question-first: Does the page answer its main question in the first two sentences, and cover the obvious follow-ups (price, alternatives, integrations, limits)? U, Unmistakable entity: Is it clear who the company is, what category it's in, and what it replaces or integrates with? Is there structured data, and does it match the visible text? O, Other people's proof: Does the page reference independent sources such as reviews, press or customer evidence? T, Truth you can check: Are prices, plans, limits and numbers stated specifically, with sources where needed? E, Extractable: Is the content split into self-contained sections with clear headings, a summary, lists or tables? D, Dated & consistent: Is there a visible last-updated date? Is anything on the page likely to contradict other sources? Finish with the three fixes that would make the biggest difference, in order. Page: [paste the URL or the page text here]
How we use QUOTED with clients
QUOTED is the standard every page we write or fix has to meet. In practice:
- In the free audit, we score your key pages against all six letters and show you which gaps to close first.
- When we write, comparison pages, alternatives pages, integration pages and pricing copy go out built to QUOTED.
- After we ship, we re-ask the assistants the same questions and check whether you're named more often, and described more accurately.
Want your pages scored against QUOTED?
It's part of every free audit. We'll show you where you stand on Google and in AI answers, and what to fix first.
Frequently asked questions
What does QUOTED stand for?
Question-first, Unmistakable entity, Other people's proof, Truth you can check, Extractable, and Dated & consistent: the six things a page needs before AI assistants will find it, trust it and quote it accurately.
Is QUOTED different from SEO?
It builds on SEO rather than replacing it. Good technical SEO already covers much of Extractable and Unmistakable entity. QUOTED adds what classic SEO tends to skip: checking what AI actually says about you, keeping your facts consistent across the web, and earning the third-party proof that shortlists are built from.
Do I need special schema or an llms.txt file for AI?
No. Google says there's no special schema, AI text file or markup needed to appear in AI Overviews or AI Mode. Structured data that matches your visible content still helps machines understand you, and llms.txt is a cheap optional extra for other assistants.
Which pages should I start with?
The pages buyers and assistants check before a purchase: pricing, the main product page, comparison and alternatives pages, integrations, security and trust pages, and docs.
How do I know if it's working?
Ask the main assistants the same set of buyer questions on a regular schedule, and track how often you're named against competitors and whether what they say is accurate. Our State Of The LLM Union tool does this weekly across ChatGPT, Claude, Gemini, Perplexity and Google's AI Overviews.
How often should I re-check?
Monthly for your key pages, and straight away after any change to pricing, plans or product names, because that's when facts drift fastest.