Google AI Overviews in Saudi Arabia: how it picks sources

Google AI Overviews is a summary Google writes itself above the search results, built from pages it read at that moment, with a few links to some of those pages beside or beneath it. It is not a search result and it is not an ad: it is a written answer the searcher reads before reaching the usual ten links.

It is not the old featured snippet either. A snippet was a verbatim quote from one page, so you could see where it came from. An overview is new text written after reading several pages, so it may carry a fact from one and a figure from another, and it may name neither. That difference is the reason for this guide: the page that wins the snippet is not necessarily the page the overview cites.

The effect in Saudi Arabia is simple. A query that used to return ten links now returns a paragraph and three links. Whoever is named in the paragraph takes the attention, and whoever is not needs the searcher to scroll past it. On a phone, which is how most of this market searches, many never scroll at all. That is why the overview is a surface to be measured rather than a technical curiosity.

It is one of eight surfaces we measure, all described on the AI engines we measure in Arabic. It reaches more of your customers than any chat app does, which is why it sits in the free plan from day one alongside ChatGPT. Getting cited in it is part of optimising for AI answers (GEO).

Google AI Mode is a conversational search surface with its own tab: a longer answer, written fresh, with references of its own. AI Overviews is the short summary above the ordinary results. Both are Google, the same question can produce two different answers with two different sets of sources, so we name them apart and measure each one separately.

• AI Overviews: above the results page, short, and it reaches everyone searching in Arabic from Saudi Arabia without asking for anything. Included in every plan, starting with the free one.
• AI Mode: behind its own tab, longer, with more clearly named sources when it names them. On Gulf Arabic questions we asked in September 2026, Google AI Mode answered every one, and some answers carried no source at all. An answer without a source has not been won by anyone yet, neither you nor your rival. The Pro and Business plans include it.

Nobody outside Google knows the rule, and any guide claiming to know it is guessing. What we can report is what we saw ourselves measuring Arabic questions in September 2026, and it is three outcomes rather than two:

• An overview with sources. The case you are looking for: a paragraph, links under it, and either you or a rival inside it.
• An overview citing nobody. Google writes a great many Arabic answers with no usable reference at all. Nobody won that question, not you and not your rival, which leaves it open to whoever builds the page worth citing.
• No answer. No overview on the page at all. We record it as an outcome in its own right, and it never counts against your site.

The overview does not search the whole web. It starts from the pages already ranking for that query and picks the ones it can quote with confidence. Three traits decide which:

• Existing ranking. If your page is not on or near the first results page for that query, it is probably outside the set being read. Classic SEO did not die; it became the entry condition.
• An explicit answer. One sentence that answers the question, carrying a number, a duration or a price, near the top of the page rather than the bottom. The overview quotes short passages, so hand it one ready-made.
• Entity clarity. Who owns the page? What is the business called? In which city? Offering what? A page that answers those in text gets quoted more than one that leaves them to the design and the photographs.

Five changes, in order, all editorial rather than technical:

1. 1Make the question a heading. Write your customer's question as a subheading, in Arabic, in their words rather than in marketing phrasing. The method is in how to choose the questions you measure on.
2. 2Answer under it in two lines. The name, the city, the figure or the duration. Then elaborate as much as you like below.
3. 3Structure the rest. A price table, a service list, a FAQ carrying real questions rather than invented ones.
4. 4Name the entity in the text. The business, the city, the district and the service written as text on the page, matching your Google Business Profile and your other pages.
5. 5Add genuine local signals. Opening hours in Saudi time, kept current through Ramadan and Eid; a Saudi address and a phone number starting +966; prices in riyals; your commercial registration number and national address where they apply; and an updated date. These help and guarantee nothing, and faking them does more harm than leaving them out.

You get what your customer saw in their own country and language, not a generic version: the overview as it was written above the results, who was cited in it, and the whole text kept with its date. Saudi Arabia and Arabic if that is your market, another country and dialect if it is not.

Because this surface moves, it is measured again on every cycle, so one lucky reading is never counted as a change, and whenever a measurement spans a range, it is shown beside the number. What you read is not one snapshot that luck picked out.

Every reading gets one of four outcomes: cited, mentioned without a link, absent, or no answer. No answer is recorded under its own name and never counts against your site, because a Google that wrote nothing said nothing about you. An overview written with no sources, on the other hand, is a real answer that you are absent from, and so is your rival, and we show that rather than hide it.

We keep the overview text and its links as we read them, dated. That is what makes every change checkable: you read the latest answer in its own words, see in “What changed” how it differs from the previous run, and follow in “How the score has moved” how your number moved over the months. It is also what lets the competitor analysis put your page beside the cited page and say why it was chosen and what is missing from yours.

And an alert reaches you only when a change is confirmed: a rival that took your place, a citation you lost, or a new gain. Whatever has not held stays in your history without an alert, so neither the numbers nor your dashboard flicker. What each run gives you is set out in how we measure AI answers.

What we measure is a picture of a surface that changes daily, not a copy of a Google database. The overview differs between countries, between languages and between hours. We hold the country, the language and the question fixed, print the date beside every number, and leave you to read the trend rather than the point.

That is why every question is measured repeatedly: a single point is not a story; four or five consecutive points moving the same way are. When a measurement spans a range, it is shown with the score, so you know how settled it is before you act on it.

We sell no guarantee of a citation, we do not know Google's internal rule, and we do not claim one number represents every question in your market. What we sell is regular measurement in Arabic on questions you choose, with the text of what was written and its sources, and an action plan ordered by priority. Plans and their limits are on pricing in Saudi riyal.