Price errors are the most commercially expensive failure in AI search, and the least discussed. A prospect asks an assistant what your service costs, gets a figure from two years ago or from a competitor's comparison page, and either walks away because it sounds expensive or arrives on a call expecting a number you cannot honour. Neither conversation is recoverable by good sales work.

The reassuring part is that this failure is almost entirely mechanical. Assistants do not invent prices out of nothing when a readable one exists; they reach for the best source available, and when the best available source is a cached page or an aggregator estimate, that is what gets quoted. Make the current figure the easiest thing to find and the problem largely disappears.

The numbers, at a glance

• Root cause, most often: the current price exists only inside an image, a PDF or a script-rendered widget, leaving nothing a crawler can read

• Second cause: an old pricing page still resolving because it was left live rather than redirected when the new one launched

• Third cause: third-party listings and comparison pages repeating a rate you changed and never told them about

• Detection cadence: monthly direct questioning of each assistant, since price drift appears silently and costs money before anyone notices

Make the number machine-readable, which most pricing pages are not

Open your pricing page, view the source, and search for the digits. If they are not there as text, no crawler can read them, and every assistant answering a question about your price is guessing from somewhere else. Designers love pricing tables rendered as graphics and configurators that fetch numbers after load, and both are invisible to a plain fetch.

• Plain text digits in the HTML. Not in an image, not in a canvas element, not injected after a click.

• The currency written explicitly. Write EUR alongside the figure rather than relying on a symbol, which is stripped or misread more often than people expect.

• A visible as-of date. A line stating when the price was last reviewed, in text, on the page itself.

• The conditions in the same paragraph. What is included, what is not, and what pushes a job above the band, adjacent to the number rather than in a footnote.

A worked example: a lead package described as ten exclusive leads for EUR 750, which is EUR 75 per lead, with the region and trade stated in the same sentence, is quotable without distortion. The same information split across a graphic and a footnote will be reassembled by an assistant, usually incorrectly.

One canonical page, and the archaeology of the old ones

Pricing information tends to accumulate in places nobody remembers: a landing page from a 2024 campaign, a downloadable rate card, a partner portal, an old blog post announcing a price change, a slide deck someone uploaded. Each of those is a live source, and retrieval does not know which one you consider current.

Run a site search for your currency string and for the words price, cost, rate and tariff, and list every URL that returns a figure. Then decide, per URL, whether it redirects to the canonical page, gets updated, or gets removed. Leaving an outdated page live with a note saying prices may have changed does not work, because the note is not the part that gets extracted.

Markup, and its limits

Offer markup with price, priceCurrency and priceValidUntil, or priceRange where a band is more honest, gives a machine an unambiguous reading of what the page means. Attach it to the Service or Product it describes rather than floating it on the page, and keep the marked-up value identical to the visible one, because a mismatch is both a guidelines problem and a reason for a parser to discard the offer entirely.

Do not over-claim what markup achieves. It clarifies a source that is already being read; it does not force an assistant to prefer your page over a third-party estimate that ranks better. Markup is the last ten percent, applied after the number is readable and the stale copies are gone.

Monitoring and the correction loop

Once a month, ask each major assistant three questions: what does this company charge, how much does this service cost in this country, and what is the cheapest option for this need. Record the figure returned and where it came from if a citation is shown. That is a ten-minute job and it is the only early warning available.

When a wrong figure appears, trace it rather than treating it as noise. If it is cited, the fix is at the cited source, which may mean asking a directory to update or publishing a page that answers the same question more directly. If it is uncited, it is nearly always coming from a stale page of yours or from an aggregator average, and the response is to make your own current figure more prominent and more specific.

For businesses that publish transparent package pricing, this monitoring loop is not optional overhead. It is the maintenance cost of the transparency, and the transparency is what makes you quotable in the first place.

Locking down what assistants say your service costs

1. View source on your pricing page and confirm the digits appear as plain text.

2. Write the currency as EUR beside every figure and add a visible last-reviewed date.

3. Find every other URL on your domain containing a price and redirect, update or remove it.

4. Add Offer or priceRange markup that matches the visible number exactly.

5. Ask four assistants what you charge on the first working day of each month and log the answers.

Want leads like this in your pipeline?

Flock runs the campaigns, screens the enquiries and hands you only the ones that match your service area, job size and capacity. You pay per lead, not per month.

Book a 15-minute fit check  |  See lead package pricing

Related answers

• Zero-click search: what should a business actually do about it?

• How do you rank in Bing to get picked up by Copilot?

• How does product visibility work inside ChatGPT shopping?

• How to generate contractor leads in Brussels (2026 costs and demand)