Pricing for AI agents, and the quote problem

Price is the field that decides whether you are comparable at all, and it is the one most small businesses deliberately withhold. That withholding does not remove you from the comparison. It removes you from the source of the number being compared.

Reviewed September 7, 2026 · 11 min read

Two things are true about price in 2026 and they pull in opposite directions. Agents compare on it, relentlessly, because it is the one attribute that is unambiguous across vendors. And a large share of businesses, especially service businesses, refuse to publish one, on the entirely reasonable grounds that the real answer depends on the job.

The resolution is uncomfortable. Withholding a price does not take you out of the comparison. It takes you out of the position of being the source. Google will reconcile a product price across four separate surfaces and disapprove you when they disagree, and for a service business with no feed at all its automated systems will simply call you and ask what things cost. Either way a number gets established. The only variable is whether it is yours.

Four places your price exists#

If you sell products, your price is not one value. It is four copies of a value, and Google checks them against each other by crawling your product landing pages and your checkout process, then comparing what it finds against your data source and your structured data.

The consequence of disagreement is more severe than most merchants expect. Products hit with a price mismatch get a preemptive disapproval, which removes them from Shopping ads and from free listings, so the penalty is not confined to the paid surface. Accounts get warned and then suspended, and once you have fixed things a review takes about seven business days, during which you are still absent.

Google crawls your checkout#

This is the part that surprises people, so it is worth stating plainly: the crawl does not stop at the product page. Google walks the checkout flow, including pre-selected variants, and the specific thing it is looking for is whether the price at checkout is higher than the price on the landing page.

That makes the classic conversion tactic a policy violation. Service fees, handling fees, carrier fees, state fees: they either go into the shipping attribute or into the base product price, and if they go into the base price they have to be visible that way on the landing page too, not itemized for the first time at checkout. The rule is not really about tax accounting. It is a no-surprises rule with a crawler behind it.

What "call for a quote" actually does#

Service businesses have no feed, so none of the above applies mechanically. What applies instead is simpler and harsher: an agent asked to compare three providers will compare the ones that gave it something to compare.

And the number gets sourced anyway. Google documents that its automated systems call businesses to verify pricing and availability, and at I/O it said Search would surface latest pricing and availability for local services with links to finish booking. So the choice was never between publishing a price and keeping it private. It is between publishing a price and having one collected from you by phone, or inferred from a directory, or simply skipped in favor of the competitor who answered.

The honest objection to publishing is that the real price genuinely varies. The answer is that a range is a price. So is a starting-at. So is a fixed fee for the diagnostic visit that establishes the real number. All three are comparable; "it depends" is not, and an agent cannot relay it to the person waiting. The phone side of this is covered in when Google’s AI calls your business.

Making a price actionable rather than decorative#

A decorative price is a number on a page. An actionable price is one an agent can put into a comparison without guessing. The difference is four things, and only the first is the number:

1. A currency and a unit. "$180" is ambiguous across per hour, per visit, per room and per month, and the unit is where service pricing loses agents most often.
2. What is included at that price, stated positively. An agent constructing a comparison needs the boundary of the offer, not a list of exclusions buried in terms.
3. What moves it, stated as a condition rather than a disclaimer. "Add $60 after 6pm and at weekends" is machine-readable in a way that "surcharges may apply" is not.
4. Structured data carrying the same number as the page. Google explicitly names unaligned structured data as a cause of mismatch, and it is the copy most likely to be stale because nobody looks at it. Templates are in the JSON-LD guide.

The freshness problem, which is the actual failure mode#

The named causes of price mismatch are mundane and they are all timing: a feed that did not update when a sale ended, prices that vary by location, structured data left behind by a page edit, a variant preselection that shows one price and charges another. None of these is a pricing strategy problem. They are synchronization problems, and they are why Google recommends the Inventory API or the Content API for anything that changes often rather than a scheduled feed, plus Automatic Item Updates to absorb small discrepancies.

Which points at the structural fix rather than the checklist. Four copies of a number, updated by four different processes on four different schedules, will disagree eventually. The businesses that do not have this problem are the ones where the four copies are generated from one source, so there is nothing to fall out of sync. Product feeds for AI agents covers the catalog side of the same argument.

What to do#

1. Buy something from yourself. Walk your own checkout on a live product and compare the final total against the landing page. This takes four minutes and it is the check nobody runs.
2. Look at your Product structured data and confirm the price in it matches the page it is on. It is the copy most likely to be silently wrong.
3. If you sell services, publish a range, a starting-at, or a fixed diagnostic fee. Pick whichever is true. Any of them makes you comparable.
4. State the conditions that move the price as conditions, with numbers, near the price itself.
5. If prices change often, move off scheduled feeds and onto an API, because the mismatch window is exactly the gap between your site updating and your feed catching up.

The thread running through all of this is that price stopped being a marketing decision and became a data integrity one. It used to be that a vague price cost you some enquiries and a surprise fee cost you some carts. Both were survivable because humans are forgiving, curious, and already halfway through. Agents are none of those things, and they are increasingly the first reader. The number has to be true in four places, or present at all, before any of the persuasion you spent money on gets a chance to work.