Do you need to publish your pricing to get recommended by AI?
No, a full rate card is not required, and a readable answer to the cost question is. No controlled study shows that hiding a price costs you citations, and we will not pretend otherwise. What is documented is that most B2B vendors publish nothing an engine can use: an audit of 134 fintech infrastructure vendors in July and August 2026 by FintechSpecs found 58.2% of them opaque, with no numeric pricing anywhere and only a quote request. The practical move is to publish a band plus what moves it, in server-rendered text, so that when an assistant answers a question about cost the best available source is yours rather than a review site's guess.
Do you need to publish your pricing to get recommended by AI?
Publishing a full price list is not a requirement for being recommended by AI, but having no readable answer to the cost question is a risk you are taking on purpose. The distinction matters because the two are usually argued as if they were the same decision.
What an assistant needs is something quotable. A band with the factors that move it, in plain text on a page it can read, gives it a sentence to repeat. A page that says pricing depends on your requirements and offers a form gives it nothing, so the answer gets built from whatever else exists: a review site estimate, a competitor's comparison table, a figure a former customer mentioned in a forum.
For considered B2B purchases the cost question arrives early and often, which is why this matters more than it used to. A buyer asking an assistant to shortlist vendors in your category is likely to ask what each one costs in the same conversation, and a vendor whose cost cannot be characterised at all is easy to drop from a list of five.
The honest version of the claim stops there. We have found no published test that isolates pricing disclosure and measures its effect on citation rates, so anyone telling you a pricing page is worth a specific percentage of visibility is extrapolating. What follows is the documented part: how few companies publish anything usable, what engines do with the gap, and what counts as readable.
What does an AI engine do when it cannot find your price?
An engine that cannot find your price does one of three things, and none of them is neutral. It states that you do not publish pricing, it repeats a third party's estimate, or it leaves you out of a comparison that includes numbers for everyone else.
The third outcome is the one to worry about, because comparison content is what gets retrieved for buying questions. Analysis of B2B pricing content by Sword and the Script in August 2025 observed generative comparison matrices including pricing as a column, and argued that vendors who omit it risk not reaching the shortlist the matrix produces. A table with a blank cell beside your name is a weaker entry than a table with a range in it.
The second outcome is quietly worse than it looks, because the number that fills the gap is rarely yours. Review profiles, aggregator pages and competitor comparison posts all carry pricing claims of varying accuracy, and an assistant with no primary source will use them. Correcting that afterwards is harder than publishing a band in the first place, and our page on why AI gets business information wrong covers how those corrections actually work.
What an engine will not do is make a number up out of nothing in the way buyers fear. Assistants with web access answer from retrieved sources, so the risk is a wrong number with a real origin rather than an invented one. That distinction is practical: it means the fix is to make your own page the most readable source on the question, not to monitor for fabrications.
How many B2B companies actually publish a usable price?
Most B2B companies publish nothing a machine can use, which is why a modest disclosure stands out more than it should. Three independent audits in different categories all land in the same territory.
The most methodical recent one covers fintech infrastructure. FintechSpecs audited 134 vendors in July and August 2026 and classified 17.2% as fully transparent, 22.4% as partially transparent, and 58.2% as opaque, meaning no numeric pricing anywhere on the site and only a contact or quote-request path.
Marketplace listings look similar. A G2 review of its own product profiles in March 2025 found only 4% of profiles explicitly listed prices, with more than 2,300 profiles carrying non-numerical pricing such as a "contact us" call to action instead. In sales tooling specifically, MarketBetter's 2026 analysis of more than 70 B2B sales tools reports roughly 3 in 10 publishing no usable pricing at all.
Read all three as indicative rather than definitive. Two are published by vendors with a commercial interest in the topic, the categories differ, and "usable" is defined differently in each. The consistent finding across them is the useful one: in most B2B categories, a page that states a band and explains what moves it is more informative than the majority of your competitors' pages, which is a cheap advantage in a system that rewards the most answerable source.
How much pricing do you actually have to publish?
How much to publish is a spectrum, and the right position on it depends on how custom your scope genuinely is rather than on how much you dislike the idea. The table sets out what each level gives an assistant and what it costs you in the sales conversation.
| What you publish | What an engine can state | Who it suits | What it costs you |
|---|---|---|---|
| Full rate card | An exact figure per tier, with the unit and billing period | Self-serve or lightly scoped products with stable tiers | Flexibility on discounting, and a visible ceiling in negotiations |
| Tiers with ranges | A band per tier and what separates the tiers | Most B2B software and productised services | Little, beyond maintaining the page when the bands move |
| Starting-from figure | A floor, and that scope above it is quoted | Services where the floor is real and the ceiling varies widely | Anchoring every conversation to the lowest number on the page |
| Cost drivers only | What changes the price, with no figure to quote | Genuinely bespoke work, such as large implementations | Being omitted from comparisons that carry a price column |
| Contact us | That you do not publish pricing | Almost nobody, as a deliberate choice | Third parties supplying the number on your behalf |
Tiers with ranges is the position most B2B companies can defend internally, because it answers the buyer's actual question, which is whether you are in their budget order of magnitude. Sales keeps the scoping conversation, the assistant gets something to quote, and nobody has published a figure that becomes a ceiling.
Where you land, write it as text rather than as a promise of a conversation. Our own figures sit on the SIGNALS pricing page, maintained by hand, which is also why you will not find them quoted inside articles here: a number copied into an article goes stale the day it changes.
What counts as a readable price to an AI crawler?
A readable price is one that exists as text in the HTML a crawler receives, with its currency, unit and billing period beside it. Four common pricing page patterns fail that test, and all four are easy to miss because they look fine to a human visitor.
Client-side rendering is the first and most common. A pricing table assembled by JavaScript after the page loads can be invisible to crawlers that fetch HTML without executing scripts, so the page an engine reads has headings and no numbers. Our guide on whether AI crawlers can read JavaScript covers which crawlers execute scripts and how to check what yours returns.
Prices inside images are the second, including the screenshot of a tier table that design produced for the launch. PDF rate cards are the third: better than nothing, worse than a page, and frequently the only place a number exists. Gated calculators are the fourth, where the figure is generated after a form submission and so never exists in any crawlable document.
The fix for all four is the same and takes an hour. State the numbers in server-rendered text, name the currency and the period, say what the figure includes, and add Offer or Product structured data so the unit is unambiguous. Our page on whether schema markup helps AI citations sets out what that markup does and does not do, and whether gated content gets cited covers the form-shaped version of the same problem.
Then check it from outside. Fetch your own pricing page the way a crawler would, with scripts disabled, and read what comes back. If the numbers are missing from that document, they are missing from the engine's view of your company, whatever the page looks like in a browser.
Can you publish a range instead of a rate card?
A range works, and for custom-scoped work a range plus the cost drivers is more accurate than any single figure would be. What makes it useful to an assistant is specificity about the band, not precision about the price.
Three elements make a range quotable. The band itself, with both ends stated. What sits at each end, so a reader can place themselves inside it. And the factors that move the number, named concretely enough to be useful: scope, volume, number of markets, how much of the work your team does rather than the supplier's.
Vague bands are worse than no band. A page saying projects range from a few thousand to several hundred thousand describes the whole market rather than your product, and an engine summarising it produces a sentence that helps nobody. If your real range is that wide, it is usually two products with different pages rather than one range.
Keep the page dated and maintained, because pricing content goes stale faster than anything else on a site. A band from 2024 still sitting on a page in 2026 is a worse outcome than the quote request it replaced, since it is now a confident wrong answer that an engine will repeat. Our guide to what AEO costs and what moves a quote is written in exactly this shape, as a worked example of describing cost without publishing a rate.
Does publishing your prices actually get you cited more?
Publishing prices has not been shown to increase citation rates in any controlled study we can find, and the honest answer to the question is that nobody knows the size of the effect. Several vendor blogs assert it confidently; none of them tested it against a comparable group of pages with the pricing removed.
What can be said with evidence has three parts. Buyer surveys consistently find pricing among the things buyers most want on a vendor site, though their numbers disagree widely enough that we would not quote a single figure. Comparison pages that carry pricing columns are retrieved for buying questions, as the Sword and the Script analysis above describes. And a page can only be quoted on a question it answers, which is the mechanism the SIGNALS model scores as alignment on our methodology page.
Putting those together gives a modest, defensible claim. A readable price makes you eligible to be cited on cost questions, which you otherwise cannot be, and eligibility is not the same as a measured lift. That is a smaller promise than the one you will read elsewhere, and it is the one the evidence supports.
If you want to know the effect on your own site, the test is straightforward enough to run. Record how assistants answer a cost question about your category and about you specifically, publish the band, then re-run the same prompts over the following weeks. Our guide on measuring AI citation decay covers the sampling discipline that makes a before-and-after like that readable rather than anecdotal, and a free visibility assessment will give you the before.
What else do people ask about pricing pages and AI search?
Will AI invent a price for your product if your pricing page has no numbers?
An assistant asked for a price it cannot find will usually say the vendor does not publish pricing, summarise a third party's estimate, or quote a figure from a review site or a competitor's comparison page. Which of those happens depends on what exists elsewhere about you, not on what you intended. The way to control it is to put a number or a range somewhere readable on your own site, so the best available source is yours.
Does a pricing page in JavaScript get read by AI crawlers?
A pricing table that only appears after JavaScript runs is often unreadable to the crawlers that feed AI answers, because several of them fetch the HTML and do not execute scripts. The same applies to prices inside images, PDF rate cards and interactive calculators. Put the numbers in the server-rendered HTML as text, then use whatever interactive layer you like on top of it.
Is a range enough, or does it have to be an exact price?
A range with the factors that move it is usually enough, and for custom-scoped work it is more accurate than a single figure. What an assistant needs is something quotable and attributable: a stated band, what sits at each end of it, and what changes the number. A page that says the price depends on your requirements without any band at all gives an engine nothing to repeat.
Will publishing prices help competitors more than buyers?
Competitors can usually find your pricing already, through review sites, former customers, lost deals and procurement documents, so the secrecy mostly costs you the buyer rather than the competitor. The genuine risk is a published figure becoming a ceiling in later negotiations, which is an argument for publishing a band and the cost drivers rather than a single rate. Where a figure would be actively misleading, say what the price depends on and how a quote is built.
Does structured data help an AI engine read your price?
Offer and Product markup makes a price machine-readable and removes ambiguity about currency, unit and billing period, which is worth doing. It is not a substitute for the visible text, because the answer an assistant writes is generated from the content it read rather than from the markup alone. Mark up the price and state it in plain words on the page, and keep the two consistent.