How do ChatGPT and Google pick an online store to recommend?

The Tasadr team · 26 September 2026 · 17 min read

Your customer asks ChatGPT and Google for the best store for a product before opening any store, and the engine names a few and cites their product pages. It recommends yours when the product page answers the question in text: price, cities shipped to, delivery time and return terms. This guide shows what to write and how to measure it.

What does a shopper ask before buying?
They ask for a product with a condition, not for a store by name. A shopper in Saudi Arabia or the wider Gulf types a sentence with the product and the condition that matters to them, such as price, delivery time or returns, and often their city. They read an answer that names a few stores with links. A store the answer does not name never enters the comparison, and its owner never learns they lost it.
So the first step is knowing your customers' questions as they type them, not as your product descriptions put them. Here are ten questions of the kind typed before a purchase, in plain English; in Arabic each is written in its city's own dialect:
- best online perfume store in Saudi Arabia
- abaya store with same-day delivery in Riyadh
- best specialty coffee store that ships to Jeddah
- genuine oud perfume under 300 riyals
- trusted phone store with instalments and no down payment
- kids' clothing store with free returns
- best Saudi store for bakhoor that ships to the UAE
- where to order dates with delivery to Kuwait
- skincare store in Saudi Arabia with genuine, registered products
- a gift delivered tonight in Jeddah with a gift card
What these questions have in common
Each one carries a condition a page in your store can answer in text: a price ceiling, a city, a delivery time, returns, authenticity, instalments. The condition is what sends the engine looking for one specific page rather than a general one, and it is what lets a small store get ahead of a large one that never wrote the condition down.
Mind the dialect, too. The same question is worded one way in Riyadh, another in Jeddah, and differently again in Dubai and Kuwait, and the engine searches with the words it was given. So each site in Tasadr is set to its own country and dialect, and questions are written the way your customer writes them there. The full method for gathering and pruning questions is in how to choose the questions you measure on.
Does the engine cite the product page or the store page?
It cites whichever page answers the question, and for buying questions that is usually a product page or a category page, not the home page. A citation is a link to one specific page, so every important page in your store is read on its own, judged on its own, and inherits nothing from what you wrote elsewhere.
The type of question decides the type of page the engine looks for:
- A product with a condition: "genuine oud perfume under 300 riyals". A product page that states the price, the size and the authenticity in text answers it.
- The best store in a category: "best online perfume store". Pages that compare several stores usually answer it, or a category page that explains the category and says who each product suits.
- Trust: "trusted phone store". Policy pages, the about page, and what other sites write about the store answer it.
- Delivery: "same-day delivery in Riyadh". A shipping page that names cities and delivery times in text answers it, with a line on the product page pointing to it.
A product page alone is not enough
The store an engine recommends has a product page that answers the product question, a category page that answers the category question, and shipping and returns pages that answer the trust question, all written as text and in Arabic. That is optimising for AI answers (GEO) for an online store: every buying question has a page that answers it in its first lines.
And what about Google?
When your customer types the question into Google, they may read an AI-written summary above the results, with links to its sources: Google AI Overviews. In Google AI Mode, Google writes a longer answer with its own references inside search itself. Both pick their sources on similar logic: a page that answers the question in its first lines. The detail is in the guide to Google AI Overviews in Saudi Arabia.
That is why store SEO remains a foundation you cannot skip: a page that never reaches the search results is rarely read by an engine at all. Ranking alone is not enough, though, because the engine picks, from the pages it read, the one that wrote the condition down.
In the measurement, a citation of any page on your store's domain is a citation of your store, and it comes with the answer's text and the links it cited, so you know which page was the source.
What do you write on a product page to get it cited?
You write, in text, what the shopper asks before paying: what the product is, what it costs, where it ships, when it arrives and how it is returned. An engine quotes a sentence it can read; it does not infer a fact from an image or from a promotional banner across the top of the store.
Seven things every important product page needs, most important first:
- The product name in your customer's words. The trade name as it is, with the description a customer would type: "long-lasting oud for evening occasions", not "Oud Intense EDP" alone.
- The price in text, as the customer pays it. A figure in riyals near the top of the page: no line asking the customer to enquire about the price, and no price hidden behind a size selector.
- Cities and delivery time. "Delivery within Riyadh in one day, to other cities in 2 to 4 days, and shipping to the Gulf countries", or whatever matches your reality.
- Returns and exchange terms in one line. How many days, in what condition, who pays the return shipping, then a link to the full policy.
- The specifications people ask about. Size, fit, material, country of origin, warranty, and whatever your customers ask in chat before they buy.
- A description you wrote yourself. A supplier's description appears word for word in many stores, which gives an engine no reason to choose yours among them. Two sentences of your own on who the product suits and when beat a copied paragraph.
- A short FAQ at the bottom. Three to five questions taken from real customer messages, each with a two-line answer.
Example: a perfume page before and after
An illustrative example, not a real store. Before the edit, the page opens with a large image and then one sentence: "A luxurious fragrance with a captivating oriental touch." The sentence is pleasant, and it answers none of the questions above.
After the edit, the description opens with two lines: "High-concentration oud perfume, 100 ml, suited to evening occasions. Price 289 SAR; delivery within Riyadh in one day; returns within 7 days if the bottle is unopened." Then a small table with size, ingredients and country of origin, then three FAQs.
Those two new lines are what an engine quotes. Every figure in them must match reality: a wrong figure is quoted as readily as a right one, and it comes back to you as an angry customer.
The store pages that back up the product page
Shipping, returns and about pages answer the trust questions, so write them for your own store rather than from a generic template: cities by name, times in days, who pays for a return, how to reach you. A category page deserves a paragraph at the top that answers the category question: how the types differ, who each suits, and where prices start.
If your customers are across the Gulf, name in text the countries you ship to and the shipping time to each, because a question typed in Kuwait or Dubai looks for a store that says so plainly.
None of this needs a developer, whether your store is on Salla or Zid: it is all text written into product descriptions and store pages.
Try this on your own site
Add your site, write 5 questions your customers ask, and see who leads the answer today. No credit card.
Start free
Why does the engine recommend a rival store instead of yours?
Usually because their page answered the question and yours did not. An engine never tries the product; it reads what was written about it: who stated the price, the city and the delivery time in text, and who left them to images and to a chat after the visit.
Six common reasons, all of them fixable:
- Their page states the condition. Price, delivery time and return terms are text on their page, and an image or nothing on yours.
- Their page was written in Arabic. Many Gulf stores write their pages in English first, with a shorter or literally translated Arabic version. An engine asked in Arabic looks for an Arabic page that answers.
- Their description was written for them. They wrote their own product descriptions; yours is the supplier's, as found in many other stores.
- Their words are your customer's words. They write "long-lasting oud for occasions" the way customers ask, and you write only the English trade name.
- Their name travels beyond their store. An engine may read a rival's name in comparison articles, lists and reviews on other sites, and recommend them because it knows them from more than one source.
- Their page can be read. Yours may be blocked to crawlers, hold text that only loads after JavaScript runs, or sit behind a pop-up that covers it.
What about large marketplaces and comparison sites?
On broad questions about a whole category, large marketplaces and comparison sites may get ahead of you, because they are built in the shape of a list. The answer is not to fight them on the broad question but to measure a narrower question, with a condition your page alone answers: a city, a delivery time, a specific type, a price ceiling.
How do you find the reason in your case?
Do not guess. Open the page that was cited instead of yours and read it the way an engine does: where is the answer, and does it carry a price, a city, a delivery time? Then put your page beside it.
That is what Tasadr does for every question: it shows who was cited instead of you, with the answer's text as written; the competitor analysis puts your page beside theirs; and the changes are ranked in an action plan. The rival in the answer may not be the competitor you know, but a smaller store that wrote a clearer page. What each run gives you is in how we measure your score. How to follow a rival in the answer month by month is in which competitor ChatGPT recommends instead of you.
How do you measure it question by question?
By asking the engines your customers' own questions, in their words and dialect, and reading for each question separately whether your store was cited and who was cited instead. The overall score sums it up, but decisions are made question by question: you can win the perfume question and lose the bakhoor question in the same week.
Every answer gets one of four outcomes: cited with a link to your store, mentioned without a link, absent, or no answer when the engine wrote none at all, which is never counted against you. Your score out of 100 is read from those outcomes.
Which address do we measure for your store?
We measure your store on its own domain, such as mystore.com, or on its own subdomain on the platform. Those two alone belong to your store, so every citation of them is a citation of you.
We do not measure the store platform's own address, or an address that is the platform's address followed by your store's name after a slash. Every store on the platform shares that address, so measuring it would credit you with every citation of any store on it, and the number would be false. That is why setup refuses the platform's address and asks for your store's own domain.
If your store has no domain of its own yet, ask your platform about connecting one in your name, so your store becomes an address that can be measured on its own. After that there is nothing to install in your store and no access to its dashboard: we measure what the engines see from outside.
Five steps for the first month
The method is one loop you repeat, walked through step by step for one engine in how ChatGPT picks which businesses to recommend. For a store it runs like this:
1. 1Add your store's domain, and set the country to Saudi Arabia and the dialect to Gulf, or to the country most of your customers buy from.
2. 2Write five questions for the category you sell most, from the ten above or from your customers' messages, or pick from what we suggest from your store's pages, or from Search Console if you connect it.
3. 3Measure before you change anything. The first run is your baseline; without it you will not know whether an improvement was yours or the engine's.
4. 4Edit one page for the most valuable question, with the two lines, the table and the FAQ from the example above. Do not edit ten pages in one day.
5. 5Measure again, and trust confirmed change only. One changed answer is not news, which is why an in-app alert reaches you only when a change is confirmed.
Which plan suits your store?
Plans are sold by questions and engines, not by products. Start on the free plan to learn where you stand, then move up once you have a decision to build on the number. Full prices and yearly billing are in Tasadr plans and prices, and which engine each plan opens is in the eight AI engines we measure.
- Free: 5 questions and one site, on ChatGPT and Google AI Overviews, measured monthly, with no credit card.
- Starter: 30 questions and one site for 499 SAR a month, adding Gemini, with a new analysis every week.
- Pro: 100 questions and 3 sites for 999 SAR a month, adding Perplexity and Google AI Mode; it suits someone running a store and a brand site, or more than one store.
- Business: 150 questions, 10 sites and all eight engines, from 4,999 SAR a month by request, with a new analysis every day.
What mistakes keep your store out of the answer?
Six common mistakes in online stores, most of them fixable in a week without a developer:
- Price, delivery time and return terms inside images and banners, not as text on the page.
- A product description copied from the supplier as it is, identical word for word to many other stores.
- Product pages in English only, or in a literal translation, for customers who ask in Arabic.
- A shipping and returns page from a generic template that names no city, no delivery time and nobody who pays for a return.
- Deleting a product page when stock runs out, instead of keeping it with a line on when it returns or what replaces it, which throws away a link an engine may have been citing.
- Judging from a single answer. An engine's answer changes from one run to the next, which is why we measure every question repeatedly, so one lucky answer is never counted as a change.
What nobody can promise you
Nobody can guarantee that an engine will recommend your store: it picks its sources afresh every time, and nobody sells a place inside its answer. What we can do is measure regularly and tell you where your store stands on each question, who took your place, and why their page is clearer.
The Tasadr score is ours, not a figure issued by OpenAI or Google, and we do not promise sales or visits: we measure your store's place inside the answer. The rest of your questions are answered in the questions people ask about Tasadr.

Frequently asked questions