On this page
Your team can spend weeks on a buying guide that sends shoppers to another store.
Most seller pages cited were stores' buying guides, built on customer reviews and tests. Even those sent shoppers to other sellers far more often than to themselves.
Keeping a guide current costs buyer time, about $21,500 a year in our example, and the larger risk is the buy link the guide sends to someone else.
Whose Pages the Shopping Answers Read
These figures are from our research, September 2026. A cited page is a page the assistant lists as a source.
A seller's own best-of page sat under 9.58% of ChatGPT's shopping answers. The same week, on the same design as our software study, a software company's own roundup sat under 47.50% of ChatGPT's software buying answers.
Google AI Mode almost never cited one: 0.62% of its September answers and 2.50% of October's. Its shopping answers rest on the product tiles Google shows, which 94.38% of its September answers and 68.75% of October's linked into. Your guide competes with those tiles for space, and the file of your products and prices you send Google is what gets products into them.
Testing and review sites lead instead: at least one of six of them sat under 39.17% of September's ChatGPT answers, four times the share of a seller's own list.
Why Store Guides Get Cited
Stores' guides were cited more than twice as often as makers' lists in September and October, and the cited guides say they lean on their customers' reviews, tests and fitting data.
A lab can test a litter. It cannot see a store's customer records or its fitting data. That is evidence a store can add, though the answers do not tell us why they cited a guide.
Where the Sale Goes
A store needs one more thing from a citation than a maker does: the answer has to send the shopper to the store to buy.
In October, five ChatGPT answers in six linked a store's or a maker's own page: product pages, category pages, store searches.
Of the 14 October answers that cited a store's guide, six named the store, often as a source of the picks, but one linked to a product of that store and seven to another seller's.
OpenAI says how that link is chosen. Merchants "are ranked based on factors like availability, price, quality, and whether they are the maker or primary seller of that item." The guide that shaped the answer is not among the factors OpenAI names.
A store that wants the sale as well as the citation needs its listing to win on those terms, which is the work on the product records ChatGPT reads, not more writing.
Only one of those fourteen answers linked the store's own listing.
The Decision Splits by What You Sell
If you run a store, write the guide, for the categories you stock deeply enough to test, and build it from evidence only you hold: your customers' reviews and returns, your staff's or members' tests, your fitting data. Say on the page how you chose.
Then make sure the listing behind each pick wins on availability and price, two of the factors OpenAI names, or the citation sends the shopper to someone else's listing.
If you make the product, a roundup with yourself first is the weakest page you can write: makers' lists were cited less than half as often as stores' guides. Publish what only the maker knows instead, the way Apollo publishes ride data: failure rates, warranty claims, measured range.
Where the Guide Picks Come From Now
In our example, a retailer of about 250 people refreshes about 12 category guides four times a year, and its two category buyers between them spend about a day on each refresh, 9% of each one's week.
Buyers and purchasing agents earned a mean $83,770 in May 2025, the Bureau of Labor Statistics reports; with benefits at 30% of private employer costs, that is about $119,700 a person.
Buying Guide Refresh ROI Calculator
What Refreshing Each Buying Guide's Picks Costs You, And What An Agent Saves
Saved Per Year
$13,160
281 hours handed back, after $3,000 a year to run it
- Hours Handed Back
- 281 hours a year
- Agent Running Cost (Estimate)
- $250 a month
- Payback Per $10,000 Of Build
- 9.1 months
On these numbers, the agent saves $13,160 a year after its own running cost, and every $10,000 of the build is earned back in 9.1 months.
On those numbers the refreshes cost $21,546 a year, and $8,387 with AI agents taking 75% of the work, our assumption ($5,387 of staff time left and $3,000 a year to run them): $13,160 and 281 hours handed back, and every $10,000 of the build earned back in 9.1 months.
The hours are the small part. The larger risk is a guide that gets cited while the buy link goes to another seller.
Reviews, returns and changing stock make a guide expensive to refresh, and the agents we build for your category buyers read those records and draft the picks for the buyer to check.
They run on the first business day of each quarter for every category with a guide, and again the day a pick's core sizes sell out. They read your own reviews, returns, sales, stock, test sheets and the guide's written criteria, and no outside site.
| The agents may | The agents never |
|---|---|
| Read your reviews, returns, stock, criteria sheet and test sheets | Open a rival's site, a tester's page or a marketplace |
| Draft the picks, a verdict line for each and a swap for any pick that sells out | Publish anything |
| Flag a pick whose core sizes run out | Hold your own brand to a softer rule than the rest |
A product is eligible when it is in stock in its core sizes, has enough reviews in the last year and stays under a return-rate ceiling; the buyer sets both numbers, and your own brand meets the same rule as everyone else's.
For each eligible product the agents score the reviews and returns against the criteria and draft the shortlist: a verdict line per pick tied to the rows behind it, and the return reason customers give most as a draft of where it falls short.
Nothing publishes itself. The category buyer approves every pick and every line, and legal reads the disclosure and any claim about your own brand.
One guide refreshed twice a year can stay with the category buyer; the agents earn their build when several categories need updates as stock and reviews change.
Start With One Category's Reviews and Returns
Bring one category's review export, returns log and criteria sheet. The first drafted picks arrive within two weeks of access, and the buyer approves each one.
What the agents may read and do is set in a written agreement, your data is processed where that agreement says, and every pick, swap and flag is logged.
Frequently asked questions
Should a brand rank its own product first on its own best-of page?
Our answers do not establish whether ranking yourself first changes the outcome. Google’s review guidance asks for firsthand evidence behind a best-product pick, so rank by published criteria and disclose that you make one of the products.
Can an AI agent write our buying guides?
The agents we build score products from your review text and returns against published criteria, then propose picks and replacements for unavailable products. Your own brand follows the same rules as competitors. A category buyer checks the evidence and approves the copy before publication.
What records do the buying-guide agents need?
Product reviews, return reasons, sales by product, current stock, and the category buyer's written criteria. Include staff or member test sheets where you have them, so the shortlist can point back to evidence your buyer can inspect.
Who approves the agents' product picks?
The category buyer approves each pick and its wording before publication. Legal checks the disclosure and claims about your own brand, which must meet the same selection rules as competing products.
Do AI shopping answers cite a store's own buying guide?
Sometimes. In our September 2026 research, a seller's own best-of page sat under 9.58% of ChatGPT's shopping answers, and 6.67% listed a store's own buying guide as a source. A software company's own roundup sat under 47.50% of software answers.
Does Google AI Mode cite sellers' own best-of pages?
Google AI Mode almost never cited sellers' own best-of pages: 0.62% of its September 2026 shopping answers and 2.50% of October's. Those figures cover sellers' pages overall, rather than store buying guides alone.
Which sites do AI shopping answers cite most?
Testing and review sites. At least one of six of them sat under 39.17% of September's ChatGPT shopping answers, four times the share of a seller's own list.
Are store guides cited more than brands' own roundups?
Yes. Stores' guides were cited more than twice as often as makers' lists in September and October, and the cited guides say they lean on customer reviews, tests and fitting data.
Does a cited buying guide send shoppers to the store that wrote it?
Rarely. Of the 14 October answers that cited a store's guide, six named the store, but one linked to a product of that store and seven to another seller's. Only one linked the store's own listing.
How does ChatGPT choose which store to link for a product?
OpenAI says merchants are ranked on factors like availability, price, quality, and whether they are the maker or primary seller of the item. The guide that shaped the answer is not among the factors OpenAI names.
Should a store write a buying guide for AI search?
For the categories it stocks deeply enough to test, built from evidence only it holds: customer reviews and returns, staff or member tests, fitting data. Then make sure the listing behind each pick wins on availability and price.
What product evidence can makers publish for AI shopping?
Makers can publish failure rates, warranty claims and measured range, evidence only they hold. The post points to Apollo's ride data as an example of publishing what the maker knows about its product.
What does refreshing buying guides cost a retailer?
The example costs $21,546 annually: about 12 guides refreshed four times a year. Two buyers together spend about a day per refresh. Assuming agents take 75% of the work, annual costs fall to $8,387, including $3,000 to run them, with 281 hours handed back.
When does a product qualify for a guide's picks?
When it is in stock in its core sizes, has enough reviews in the last year and stays under a return-rate ceiling. The buyer sets both numbers, and your own brand meets the same rule as everyone else's.
When do buying-guide agents run and refresh picks?
They run on the first business day of each quarter for every category with a guide. They run again the day a pick's core sizes sell out, drafting replacements for the category buyer to approve.
Are buying-guide agents worth it for one guide?
One guide refreshed twice a year can stay with the category buyer. The agents earn their build when several categories need updates as stock and reviews change, rather than on that schedule for one guide.
Can buying-guide agents read competitors' websites?
No. Buying-guide agents read the retailer's own records and written criteria, with no outside sites. They never open a rival's site, a tester's page or a marketplace to gather evidence for the guide.
How do I start with buying-guide agents?
Bring one category's review export, returns log and criteria sheet. The first drafted picks arrive within two weeks of access, and every pick, swap and flag is logged.
