What Does AI Tell Buyers About Pricing You Have Not Published?

Pricing visibility is the degree to which an AI answer about what your product costs is sourced from material you published and control.

Analysis of Profound's dataset of AI conversations found that the share of ChatGPT queries carrying commercial intent rose from 13.9% to 19.2% over a year, and that prompts containing cost or pricing keywords now represent between one and two per cent of all LLM queries. One per cent of a very large denominator is a substantial volume of buyers asking, in effect, "what does this cost and is it fair?"

That question is now frequently asked of an assistant that has read the pricing page, the review platforms, the procurement databases and the complaint threads, and will synthesise all of them into two paragraphs with a tone attached.

The tone is the part most teams have not priced in. In the study, researchers ran "evaluate company X on pricing" across six engines (ChatGPT, Google AI Mode, Google AI Overviews, Gemini, Copilot and Perplexity) against the Cloud 100 list of leading private cloud companies, collecting 7,600 responses between 29 July and 10 August 2026. For one well-known software company, negative pricing language appeared in 72 of 76 responses. For a vertical AI company with no public pricing page, negative language appeared in 67 of 76, and the engines quoted a range spanning roughly twenty-four times from bottom to top for the same product, assembled largely from forum threads.

A buyer reading that arrives at the first call either frightened off or holding a number you never published as a negotiating anchor. Forum-sourced figures are one reason third-party mentions shape what AI says about a brand.

A separate July 2026 study crawled 304 reachable B2B SaaS companies, then asked a leading model what each product cost. Among the 277 with a pricing page, 24% showed no prices at all, only contact-sales walls. For those hidden-pricing vendors, the model produced concrete dollar figures in 93.9% of answers, against 95% for the public-pricing control group. Only about 3% of answers stated that pricing was not public and stopped there.

The behaviour is worth understanding precisely, because it explains why the usual objection misses. The model knows the pricing is not published. It says so. Then it estimates anyway.

The same study found that only 19% of B2B SaaS pricing pages carry machine-readable Offer schema, so even among vendors who do publish, the number is mostly presented in a form a machine has to infer rather than read. Structured data and entity identity are the foundation for declaring a figure rather than hoping it is inferred.

This reframes the strategic question. "Should we publish pricing?" is a commercial decision with legitimate arguments on both sides. "Should the internet's answer about our pricing be sourced from us or from a forum?" is not really a decision at all.

Across the 7,675 responses in the Cloud 100 study, company-owned pricing pages appeared somewhere in 46% of answers but appeared first in only 12%. No company in the set had its own pricing page cited first in a majority of runs. Engine behaviour varied: ChatGPT defaulted to company-owned pricing pages first 38% of the time, while Google AI Mode, AI Overviews and Gemini tended to bury them.

Three external domains recurred across the dataset: Vendr, a pricing-data and negotiation provider, cited in 18.7% of runs; Reddit, cited in 18.6%; and G2, in 15.9%.

Source: Growth Unhinged and Profound, 7,600 AI responses across six engines, 29 July to 10 August 2026. Engines change citation behaviour frequently, so this is a dated snapshot.

The uncomfortable implication is that one of the three recurring external sources is a review platform where vendors can legitimately maintain accurate information, one is a procurement dataset, and one is a public forum where correction requires outreach rather than editing. How Reddit and forums feed AI brand mentions is its own discipline.

A machine-readable pricing page is one whose plans and figures appear in the initial HTML response, so a crawler can read them without running scripts or clicking.

In the Cloud 100 study, 57 of 77 public pricing pages were fully readable to bots, leaving roughly a quarter that were not. Ten pages hid at least 40% of their body content from crawlers. The researchers attributed this to four recurring causes:

• Client-side rendering. Pricing tables that populate only after JavaScript executes. Crawlers with limited or no JavaScript execution receive an empty shell.
• Interactive elements. Tabs, accordions and price calculators. Content that exists only after a click does not exist for a crawler.
• Robots.txt and infrastructure blocks. Sometimes deliberate, often inherited and never revisited.
• iFrames. Pricing widgets served from a separate, uncrawlable domain.

A parallel audit of B2B SaaS domains reached the same conclusion about the mechanism, noting that the major AI crawlers retrieve raw HTML rather than executing JavaScript, so a React or Vue pricing component that fetches its data client-side presents empty containers. Google's own guidance points the same way: important content should be available in textual form and structured data should match the visible text. The wider technical AEO checklist covers rendering and structure beyond pricing.

There is at least one documented before-and-after. Profound found its own pricing content partially hidden from bots, moved the rendering server-side, and deployed the fix on 25 June 2026. Its pricing page became the second most-cited page on the site, and citation bot traffic rose 13% week over week. That is one company's result, self-reported, with no control, but the mechanism is straightforward enough that the direction is credible.

A pricing answer stack is the set of first-party pricing, FAQ, supporting and billing-documentation pages that together give AI engines an authoritative source for what a product costs.

The strongest performers in the Cloud 100 study were not simply the most transparent. Plaid led with its pricing page cited first 42% of the time, followed by Fireworks AI, Fal AI and ElevenLabs. What distinguished them was breadth of first-party pricing material rather than a single well-written page.

For Plaid, AI answers cited its billing documentation in 70% of answers, its pricing page in 64%, and its FAQs in 50%, with several Plaid-owned pages frequently appearing in one response. The documentation, not the pricing page, was the most-cited pricing surface, because documentation is denser: it defines billing models at the product and endpoint level, covers costly edge cases, and segments by plan. Plaid's documentation also states plainly where figures are unavailable, giving engines an authoritative explanation for the absence rather than leaving a gap for a forum to fill.

Five components recur across the strong performers:

1. One clear pricing page with plans and, where possible, figures.
2. FAQs phrased as the questions buyers actually ask.
3. Supporting content on value, operational implications and edge cases.
4. Technical billing documentation covering billable events, minimums, commitments, overages, cancellation behaviour and regional premiums.
5. AI-readable infrastructure underneath all four.

It is also worth noting what the data does not support. Several companies in the study sat well above the median on citation performance despite showing plans without price points. Full transparency is not the prerequisite; readability and coverage are. A vendor with genuine reasons to keep figures confidential can still own the answer by documenting the billing mechanics, the variables that drive cost, and a concrete entry point such as a trial or a starting package. For a view from the buy side, see what GEO services cost.

Answer engine optimisation (AEO) is the practice of structuring content so AI assistants retrieve it and cite it as the answer to a buyer's question.

The work is unglamorous: getting those teams to agree on what is publishable, rendering it server-side, mirroring it in documentation and FAQs, and then keeping the third-party record accurate as plans change. Lifewood's teams tend to approach it as a consistency problem across a brand's surfaces and markets, the same discipline that applies to product data, rather than as a page rewrite. The AEO service page and the GEO service page describe how that work is organised.

Useful work here comes from several directions. Profound supplies the conversation-level data behind the Cloud 100 study. Procurement platforms such as Vendr and review platforms such as G2 hold much of the external record, and engaging with them is often higher-yield than another owned page. Independent researchers published the pricing-blackout crawl that quantified the hidden-pricing behaviour. A brand serious about this generally needs the measurement, the remediation and the off-site correction running together.

• Establish a baseline. Run "evaluate [company] on pricing" through several engines daily for a week. Record the figures, the sentiment and every cited source.
• Curl your pricing page. Fetch it with JavaScript disabled. If the numbers are absent, that is your first ticket and it belongs to engineering, not marketing.
• Move pricing to server-side rendering. Prices should appear in the initial HTML response.
• Add Offer schema so the figure is declared rather than inferred.
• Publish the billing mechanics even if you withhold the figures. Billable events, minimums, overages, commitments, cancellation terms and regional differences are highly citable and rarely confidential.
• Give one concrete entry point. A trial, a free tier, or a starting-at package anchors the range.
• Explain absences explicitly. A sentence stating why figures are quoted rather than published gives engines something authoritative to cite instead of a guess.
• Correct the external record. Update what you control on review platforms; pursue outreach for procurement databases and forum threads.
• Date everything. Current plan definitions, visible update dates and scheduled changes make an owned source safer to cite.

Teams comparing outside help can start from a ranked list of the best AEO and GEO agencies.