Our pricing page never shows up in answers

What is actually happening

You open ChatGPT. You type something a buyer would type. "How much does project tracking software cost for a team of ten?" The answer names four products with numbers next to them. None of them is yours. Then you ask again and it cites a roundup blog from last year that got your price wrong.

This is not a ranking problem in the old sense. It is a quoting problem. An assistant builds an answer out of short passages it can lift and attribute. If your pricing page holds the number in a styled table cell, a tooltip, a toggle between monthly and annual, or an image, there is no sentence to lift. A page can be indexed, fast and well linked and still give the model nothing it can safely repeat.

The six causes, in the order we see them matter

• No quotable sentence. The page shows 29 in a big font with "per user" underneath in grey. There is no line that reads "Pro costs 29 US dollars per user per month, billed annually, excluding tax."
• Price is hidden behind an action. A "contact sales" wall, a currency picker that loads on click, or a calculator. If a human needs two clicks, a crawler usually gets nothing.
• Price renders in JavaScript after load. Some crawlers do not execute scripts. The HTML they receive has an empty price slot.
• The crawler is blocked. AI crawlers use their own user agent names. A robots.txt rule written in 2022 can block them without anyone noticing.
• Nobody else repeats your number. Models lean on corroboration. One page saying 29 dollars is weaker than your page plus a directory plus a review site that all say 29 dollars.
• The page does not answer the question that was asked. Buyers ask about a team size, a use case or a comparison. A bare plan grid answers none of those in words.

Google states in its own documentation that there is no special markup that makes a page appear in AI features, and that standard Search guidance applies. See Google Search Central: AI features and your website. So the work is not a trick. It is making the number readable, reachable and repeated.

How to confirm it is happening to you

Do not ask an assistant "what does [your brand] cost?" Naming your business primes the model. It will go and find you, and you learn nothing about whether a buyer would have found you. Ask the question a stranger would ask.

Ten steps, in order

1. Write 20 questions a buyer would type before they know you exist. Include a team size, a budget, a country and a competitor name in at least five of them. Example: "Cheapest invoicing tool for a two person agency in India under 1000 rupees a month."
2. Open a fresh session with no memory and no custom instructions. In ChatGPT, turn off memory and use a temporary chat. Logged in accounts carry history, and history changes answers.
3. Ask all 20 questions in ChatGPT. Record, for each: were you mentioned, was a URL of yours cited, was a price stated, and was it correct.
4. Repeat the same 20 in Gemini, Perplexity, Claude and Google AI Overviews. Same day. The same question gets different sources from different apps, so one app is not a reading.
5. Count. Twenty questions across five apps is 100 answers. That is your denominator.
6. Fetch your own pricing page the way a crawler does, with scripts off: curl -A "OAI-SearchBot" https://example.com/pricing. Then search the returned HTML for the number. If it is not there, no model can read it.
7. Open https://example.com/robots.txt and search for the AI crawler names in the table below. Check for blanket Disallow: / under any of them.
8. Check indexing. In Google Search Console, run the URL Inspection tool on the pricing page and confirm it is indexed and that the rendered HTML contains the price text.
9. Search the open web for your price as a string, in quotes, with your brand name. Count how many sites other than yours state it.
10. Save all of it in a spreadsheet with the date. Without a dated baseline you cannot tell a fix from a mood swing.

Crawler names to look for, by vendor documentation

Last verified: 18 October 2026

Crawler names change. Open the vendor page before you edit robots.txt, not after.

What to change, in order

Cheapest first. Do not start with a site rebuild.

1. Write the price as a sentence. One line per plan, in body text, above or beside the table. Include the number, the currency, the unit, the billing period and the tax treatment. "Pro is 29 US dollars per user per month when billed annually. Tax is added at checkout." Date it: "Prices as of 18 October 2026."
2. Render the price in server HTML. If your currency switcher or monthly to annual toggle is client side, make one state the default in the HTML. Re-run the curl check in step 6 above.
3. Unblock the crawlers you want. Decide per vendor. If you want to be quoted in ChatGPT answers, OAI-SearchBot must be allowed. Blocking the training crawler and allowing the search crawler is a legitimate position and the docs explain how.
4. Add Offer structured data. Schema.org Offer expects price and priceCurrency. Google's product structured data guidance sets out the fields. Markup does not force a citation. It removes ambiguity about which number is the price.
5. Answer the real questions on the page. Add short headed blocks: "What does it cost for a team of ten?", "Is there a free tier?", "What is not included?", "How does it compare on price to [category alternative]?" Answer each in two to four sentences with the number repeated in words.
6. State what is excluded. Overage, onboarding fees, minimum seats, annual lock. Assistants quote caveats, and a page that carries its own caveats gets quoted more often than one that hides them.
7. Get the number repeated off site. Your own page is one source. Category directories, your own docs, a changelog post, comparison pages that will accept a correction, and any partner listing. Ask for the exact sentence, not a paraphrase.
8. Keep the URL stable. If you must move the page, redirect with 301 and keep the old slug alive. A changed URL resets whatever corroboration you had.
9. Re-measure at 30 days with the same 20 questions. Same wording, same no-memory sessions, same five apps.

Worked example you can repeat

A small invoicing tool runs 20 blind questions across 5 apps on 18 October 2026. That is 100 answers. Their pricing page is cited in 7 of them.

• Raw rate: 7 of 100, which is 7 percent.
• 95 percent range (Wilson interval): about 3 percent to 13 percent.

They do steps 1 to 7. Thirty days later they run the same 100 questions. Now the page is cited in 11.

• Raw rate: 11 of 100, which is 11 percent.
• 95 percent range: about 5 percent to 18 percent.

Three to thirteen overlaps five to eighteen. So the honest report is: no measured change yet. That reads like bad news. It is not. It is the difference between a real result and a story. If they had reported "citations up 57 percent" they would have been describing noise. To separate 7 percent from 11 percent with confidence you need a bigger denominator, which means more questions, more apps or more rounds. Run the same set monthly and the trend becomes readable even when any single month is not.

How long before it moves

There are three clocks and they run at different speeds.

• Crawl and index. Your edited page has to be fetched again. You can request this for Google Search through URL Inspection in Search Console. Other crawlers come on their own schedule.
• Retrieval. Assistants that search live, such as Perplexity and ChatGPT search, can reflect a changed page soon after it is re-fetched. Assistants answering from model memory alone will not, until the next model update.
• Corroboration. Getting a third party to publish your number takes as long as that third party takes. This is usually the slowest clock and the one with the biggest effect.

We do not publish a typical number of days, because we have not measured one we would stand behind. What we do instead is fix the measurement cadence: re-read every 30 days, same questions, same apps, and report ranges rather than point numbers. If two readings overlap, nothing moved yet.

One caution. A pricing page that starts getting cited can still get cited wrongly, with an old number from a cached roundup. Keep a dated "prices as of" line and keep correcting the off site copies. If assistants do not seem to know your business exists at all, the pricing page is the wrong place to start: see #. If the problem is broader than pricing, see /blog/our-biggest-page-is-never-cited.

Where AI Knows Us fits

AI Knows Us measures across ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. Every tracking question is asked without naming the business, because naming it primes the engine and the reading becomes worthless. Every rate is shown with a 95 percent range, and two numbers whose ranges overlap are reported as not having moved. The product also does the fixing, not only the measuring: it writes the pages, publishes them, and asks other sites to name the business. It runs 13 agents to do that work: intake, discovery, analyst, diagnosis, planner, writer, editor, publisher, scout, outreach, experiment, lifecycle and the copilot. It is made by Clyra Labs in India.

Paid plans are not open yet. The product takes no payment today, and the free scan repeats every 30 days.

Other tools in this category

Several products track how assistants answer. We have not benchmarked them against ours, so we are not ranking them here. Check their own pages for current features and prices:

• Profound
• Peec AI
• Otterly.AI
• Scrunch AI

If what you need today is a dashboard plugged into a large existing marketing stack, one of those may suit you better. Our own strength is the blind question method and the range reporting, plus the fact that the same system writes and publishes the fixes.

Who should not choose AI Knows Us

• You need a signed enterprise contract and a paid SLA today. Paid plans are not open yet.
• You want a single headline percentage with no range. We will not print one, because a reading off a handful of questions is not a percentage.
• You want brand-named prompts tracked ("what does [brand] cost?"). We ask blind questions on purpose, so that metric is not what we produce.
• Your pricing is bespoke and never published. If there is no number on the page, no tool can get a number quoted.
• You want daily readings. A 30 day cadence with ranges is what the method supports.

FAQ