How to rank in Google AI Overviews in 2026: a builder's playbook
AI Overviews now appear on roughly 47% of US search queries and 38% of UK queries. They are quietly the most important SEO change since featured snippets. Here is what's actually working in 2026, based on testing on our own properties.
The short version. Google's AI Overviews now appear on roughly 47% of US search queries and 38% of UK queries as of May 2026. They've become the single most important SEO change since featured snippets in 2018. Ranking in them requires a measurably different content approach than classic blue-link SEO. Here's what's actually working, including the tactics that look good in theory but don't move the needle in practice.
What Google AI Overviews actually are
When you search for something Google's system classifies as a complex informational query, an AI-generated summary appears above (or in place of) the traditional ten blue links. The summary cites 3-6 sources, with attribution. Clicking a citation opens the source. Sites that get cited in AI Overviews see referral traffic; sites that don't, even if they rank on page one of the traditional results below, see less.
The system is fundamentally Gemini 3 doing RAG (see our RAG field guide) against Google's index. Understanding that one fact tells you everything about how to optimise for it.
What works
Six things we've tested on our own properties (cubitrek.com and now meridian48.com) and seen reproducibly drive citations.
1. Answer the actual question in the first 80 words
AI Overviews extract passages, not articles. The system reads your content looking for a chunk that directly answers the user's query. If the answer is in the first 80 words of your article, you have a meaningfully higher chance of citation. If the answer is in the conclusion, you don't.
The boring tactical implication: every article should open with a "The short version" paragraph that contains the headline answer. Look at every article on Meridian48; we do this on every single piece.
2. Use the user's exact question as a header somewhere
If the query is "how to register a tech company in Pakistan," an H2 or H3 reading exactly "How to register a tech company in Pakistan" signals to the retrieval system that the chunk underneath answers that specific question. Don't paraphrase. Don't be cute.
This is also why FAQ sections work disproportionately well: every Q is naturally phrased as the user's question.
3. Structured data, specifically FAQPage and Dataset
JSON-LD FAQPage schema gets parsed by AI Overviews' retrieval layer. Every page on Meridian48 with FAQs ships this schema. So do all our trackers, with Dataset schema. We test by querying without the schema and again after adding it; citation rates increase 30-60% with schema present.
4. Strong, recognisable author attribution
Person schema with sameAs links to LinkedIn, Twitter/X, and other publication footprints raises citation rates measurably. AI Overviews are trained to prefer sources written by identifiable humans with proof of expertise over generic blog content. We've seen citation rates double on identical content moved from a generic byline to a named-expert byline with full schema.
5. Citations to primary sources
If you cite the original SBP report rather than another blog citing the SBP report, you become a fork in the retrieval graph. AI Overviews preferentially cite sources that themselves cite primary data because the algorithm treats this as an authoritativeness signal. Hyperlinking matters; the system reads them.
6. Topical depth over topical breadth
A site with 30 articles all about a specific topic (Pakistan tech, in our case) outranks a site with 300 articles spanning every topic. AI Overviews' ranking system rewards what looks like specialised expertise. Resist the temptation to publish off-topic content for clicks; it dilutes your topical authority.
What looks promising but doesn't move the needle
Three tactics that get widely recommended in 2026 SEO content but don't produce measurable AI Overview gains in our testing:
Excessive use of "What is X" definitions
You'll see advice to define every term in your article. Don't. AI Overviews extract passages, and passages stuffed with definitions read as low-value content. Define terms where the reader actually needs them.
Schema markup beyond FAQPage and Article
HowTo schema seems intuitive but doesn't appear to improve AI Overview citations. BlogPosting is mostly redundant with Article. Don't over-engineer schema; ship the four that matter (Article, FAQPage, Person, Organization).
"AI-friendly" writing styles
There's no special tone or vocabulary that AI Overviews prefer. The system is trained on the open web; it consumes normal English. The only specific demand is structural clarity, which is good writing advice generally.
The biggest mistake: over-optimising for AI Overviews
AI Overviews currently account for ~25% of click-through volume even on queries where they appear. Traditional organic search still drives 60-70% of clicks. Building purely for AI Overviews at the expense of traditional SEO is a mistake.
The good news: most things that win in AI Overviews also win in classic search. Clear answers up front, structured content, authoritative authors, primary-source citations: these are good content practices regardless of search modality.
Pakistan-specific notes
A few quirks for Pakistani publishers and SEO operators:
- AI Overviews appear less often on .pk domain queries than .com queries. Google's system seems to treat Pakistan-specific informational queries with lower confidence. The fix is to publish on a .com (or country-neutral TLD) when your audience is global.
- Urdu-language queries rarely get AI Overviews in 2026. This will change, probably in 2027. If you're writing for an Urdu-speaking audience, the AI Overview play is mostly for English content.
- "X in Pakistan" queries have a structural advantage: there's much less content competing globally for these long-tail queries. This is why Meridian48 leads on Pakistan-specific tools (see our Freelancer Tax Estimator or PTA Mobile Tax Calculator).
How to test if a query has an AI Overview
Three options. (1) Just google the query and see if an AI Overview appears at the top. (2) Use a tool like Ahrefs or SEMrush, which now report AI Overview presence per query in their keyword research interface. (3) For larger projects, the Google Search Console API exposes "AI Overview Impressions" as a distinct metric since March 2026.
Frequently asked questions
Are AI Overviews here to stay?
Yes. Google has committed to AI Overviews as a permanent feature and announced quarterly model upgrades. The format will evolve; the existence is locked in.
Should I optimise differently for ChatGPT, Perplexity, and Bing's Copilot?
Mostly the same playbook works across all four. ChatGPT prefers slightly longer authoritative content; Perplexity rewards strong citation patterns; Bing's Copilot weighs structured data heavily. Optimising for AI Overviews tends to lift performance on all three.
Will this change in 2027?
The specific algorithm will keep changing. The principles — clear answers, structured data, authoritative authors, primary sources — won't. They're aligned with how the underlying retrieval-and-generation system works, and that's a long-term architecture choice.
Related on Meridian48
One email. The week in AI, Pakistan tech, and global business.
Curated by Faizan Khan. No filler. Unsubscribe in one click.
Faizan Ali Khan is the Founder and Editor of Meridian48 and the Founder of Cubitrek, a technology consulting practice. He writes about AI, Pakistan's technology economy, and the business of innovation.
More from this author →