Google AI Overviews choose their sources in two stages: the system breaks a query into sub-questions and retrieves candidate passages for each, mostly from pages already ranking well, then the model composes an answer and cites the passages it could lift most cleanly. Analyses consistently find heavy overlap between AI Overview citations and the top organic results, but the unit of competition has changed: you are no longer ranking a page against pages, you are offering passages against passages. The pages that get cited answer a specific sub-question directly, in extractable form, on a site Google already trusts.
The stakes stopped being theoretical a while ago. By early 2026, AI Overviews were appearing on roughly 48% of Google searches and reaching around two billion people a month. When an Overview appears and you are not in it, studies put the organic click-through drop at up to 61%. When you are cited, the same research shows clicks running about 35% higher. Same query, same position, opposite outcomes. Being a source is the whole game now, so here is how the choosing actually works.
The mechanism: query fan-out and passage retrieval
When someone searches a question with any complexity, the system does not process it as one query. It fans out: “best CRM for a small business in Qatar” quietly becomes several sub-questions about pricing, Arabic support, local integrations, and alternatives. For each strand, Google retrieves candidate passages, sections of pages, not whole pages, ranks them for relevance and reliability, and hands them to the model. The model writes the Overview and attaches citations to the passages that supplied each claim. Two practical consequences follow. First, you can be cited for one excellent section of an otherwise ordinary page. Second, a page that covers the sub-questions your market actually asks has several lottery tickets in every draw, while a page that only chases the head keyword has one.
What the citation data keeps showing
Independent analyses of millions of AI Overview citations agree on the broad strokes. Ranking still matters: a large share of cited pages sit in the top ten organic results, so classic SEO remains the entry fee. But the overlap is far from total, which is the opportunity: pages outside the top three routinely get cited over pages above them because their passages answer more directly. Freshness bias is real, with recently updated content over-represented. And corroboration counts: claims that appear consistently across independent sources get synthesised confidently, while contradicted or uncorroborated claims tend to get dropped. None of this should surprise readers of my GEO explainer: AI Overviews are the same retrieval logic as ChatGPT and Perplexity, run by the company that already indexes your site.
Becoming a source: the work in order
Hold your organic ground
Since citations draw heavily from ranking pages, the boring foundation stands: technical health, real authority, content that deserves its position. If you are nowhere in the top twenty, fix that first, and my honest SEO timeline applies unchanged.
Map the sub-questions, then answer each one cleanly
Take your money query and list what the fan-out will ask: costs, comparisons, timelines, risks, local specifics. Give each sub-question its own section with a descriptive heading and open that section with a two-to-three sentence direct answer before elaborating. That opening block is the passage the system lifts. Every article on this site opens exactly that way, deliberately.
Structure so machines stop guessing
Short paragraphs, one idea each. Real tables for genuinely tabular comparisons. FAQ sections phrased the way people ask. Article and FAQPage schema declared rather than implied. This is not decoration; it is the difference between a passage that can be quoted verbatim and one that needs interpretation, and retrieval systems prefer not to interpret.
Keep cornerstone pages demonstrably alive
Given the freshness bias, an annual honest revision of your important pages, updated numbers, revised claims, corrected dates, keeps them in the candidate pool. A page frozen since 2023 is quietly aging out of eligibility regardless of quality.
Say the same thing everywhere
Corroboration is a ranking force inside the Overview. Your claims, your positioning and your basic facts should match across your site, LinkedIn, directories and press mentions, the same entity discipline I detailed in how to get cited by ChatGPT. The engines differ; the trust math does not.
What not to do
Do not block Google’s AI systems in a fit of principle unless you are a publisher whose business is the content itself; for a business that wants customers, opting out of Overviews is opting out of the answer. Do not stuff pages with question headings that repeat the keyword without adding information, since passage retrieval rewards answers, not echoes. And do not chase every “AIO optimization” tool arriving weekly; the durable playbook above is unglamorous and works across every answer engine at once.
The Arabic gap, again, and bigger
AI Overviews increasingly serve Arabic queries, and the candidate pool there is thin in most Qatari sectors. The same page quality that fights for citation among dozens of English competitors can be nearly alone in Arabic. If you have been reading this series, you know where I stand on Arabic search: it is the cheapest authority on the market, and AI Overviews just raised what that authority pays.
Frequently asked questions
How do I know if AI Overviews show for my keywords?
Search your money queries and look, in English and Arabic, logged out and in an incognito window. Patterns emerge fast: informational and comparison queries trigger Overviews far more than pure navigational or transactional ones. Track which of your target queries show one and who gets cited; that list is your competitive map.
Can I measure traffic from AI Overviews?
Imperfectly. Google folds Overview impressions and clicks into normal Search Console data without a separate label, so watch proxies: impression spikes with falling CTR suggest an Overview appeared above you; rising brand searches and direct traffic often follow being cited. It is fuzzier than anyone would like, and still worth watching monthly.
Does schema guarantee a citation?
No. Schema removes ambiguity about what your content is, which helps eligibility; the citation is earned by the passage itself. Think of structured data as clean handwriting on the exam: it will not supply the right answer, but it makes sure the right answer gets read.
Are AI Overviews killing SEO?
They are redistributing it. Zero-click behaviour is real and clicks per query are falling, but cited sources take a larger share of the clicks that remain, plus a visibility dividend inside the answer itself. SEO is not dying; the trophy moved from position one to citation one.
Is optimizing for AI Overviews different from optimizing for ChatGPT?
Eighty percent identical: answer-first passages, structure, entity consistency, corroboration, freshness. The main difference is that Overviews lean harder on your existing Google rankings, while chatbots range wider across the open web. Do the shared work once and both engines pay you.
Get into the answer before your competitors do
Most Qatar businesses have not noticed that the result page changed underneath them. That is the window. See how I build AI visibility and GEO programmes in Qatar, start from the GEO foundations if you are new to this, or book a free call and I will show you which of your money queries already carry an AI Overview, and who is being cited in it instead of you.