How to Rank in AI Overviews When the Studies Disagree

TL;DR

• Ahrefs studied 1.9 million citations from 1 million AI Overviews and found 76.1% of cited pages rank in Google's top 10, with the top-cited URL sitting at a median organic position of 2 (Ahrefs, July 2025).
• BrightEdge reported that only 17% of AI Overview citations come from top-10 pages (BrightEdge, February 2026). The two studies measure different citation slots, which is why both can be right.
• 99.2% of keywords that trigger an AI Overview are informational in intent (Ahrefs, 2025). Most of a B2B site's commercial keyword set never triggers one, so most of it is out of scope.
• The presence of an AI Overview correlated with a 34.5% lower position-one CTR in Ahrefs' April 2025 study, and 58% in its February 2026 re-run on December 2025 data.
• The typical AI Overview in Pew's sample was 67 words long and cited three or more sources 88% of the time (Pew Research Center, July 2025). That is the size of the hole your answer has to fit.

How do you get featured in Google AI Overviews?

Rank in the top five organic results for an informational query, then place a self-contained answer of roughly 40 to 70 words directly under a heading that matches the query wording. Google's documentation states there are no additional requirements to appear in AI Overviews and no special optimizations necessary, beyond a page being indexed and eligible to appear with a snippet.

Two mechanics explain why that works. Google describes retrieval-augmented generation, where its core ranking systems retrieve pages from the Search index and the model generates a grounded response from them. It also describes query fan-out, where the model issues several related sub-queries around the original question and pulls from all of them.

So the AI Overview is not a separate index with separate rules. It is a synthesis layer sitting on the ranking system you already work with. The unit it selects is a passage, not a page.

Do you need to rank in the top 10 to be cited?

For the visible citation slots, yes, almost always. For the full citation set, no. This is the question the two headline studies appear to answer differently, and the difference is methodological rather than factual.

Ahrefs analysed 1.9 million citations from 1 million AI Overviews, looking at the top three most visible citations in each response. It found 76.1% of cited pages rank in the top 10, 9.5% rank between positions 11 and 100, and 14.4% do not rank in the top 100 at all. The median organic position of the top-cited URL was 2.

76.1%: share of AI Overview cited pages ranking in Google's top 10, across 1.9 million citations (Ahrefs, July 2025)

The citation slot tracks rank tightly:

BrightEdge, tracking AI Overview citations across nine industries, reported in February 2026 that only 17% of citations come from pages ranking in the organic top 10.

Both can hold. Ahrefs restricted its sample to the three most prominent citations per Overview. BrightEdge counted across the full citation set, and AI Overviews frequently cite well beyond three sources. Pew found that 88% of the AI summaries in its sample cited three or more sources. The prominent slots correlate strongly with organic rank; the long tail behind them does not.

Treat this as a stated hypothesis rather than a settled finding. Neither publisher has released the other's raw data, and no study has yet measured citation slot position against rank across a full citation set. It is the most plausible reconciliation of two credible datasets, and it should be labelled as such.

What it means in practice: compete for the visible slots on queries where you already rank in the top five. Do not spend on pages ranking 20th in the hope of catching a long-tail citation that nobody scrolls to. Ahrefs' own Si Quan Ong put the ceiling honestly, noting that ranking first makes citation more likely than ranking lower, but that the chance is close to a coin flip.

Which of your keywords can even trigger an AI Overview?

Mostly your informational ones. Ahrefs found that 99.2% of keywords that trigger AI Overviews are informational in intent. For a B2B company, that single figure removes most commercial and transactional keywords from scope before any work begins.

Coverage also varies sharply by sector. BrightEdge reported AI Overview presence at 88% in healthcare, 83% in education and 82% in B2B technology as of February 2026, against roughly 48% across all tracked queries.

Run this check before budgeting anything. In Ahrefs, open Site Explorer, go to the Organic keywords report, and apply the SERP features filter set to AI Overview. That returns exactly which of your ranking keywords currently trigger the feature. Split the result three ways:

1. Triggers an AI Overview and you rank 1 to 5. In scope. This is where the work pays.
2. Triggers an AI Overview and you rank 6 or worse. Out of scope for now. Fix the ranking first; the citation follows rank, not the other way round.
3. Does not trigger an AI Overview. Out of scope entirely. If these are your commercial keywords, that is normal and it is good news, because those pages keep their clicks.

Most B2B sites find that group one is far smaller than expected. That is the finding that should set the budget.

How to rank in AI Overviews: the five steps

Time required: about 90 minutes for the audit, then 30 to 45 minutes per page. Prerequisites: Google Search Console verified, a rank-tracking tool with a SERP features filter, and a list of pages already ranking in positions 1 to 5.

Step 1. Confirm the page is eligible at all

Google requires a page to be indexed and eligible to appear with a snippet. Check the page source and HTTP headers for noindex, nosnippet, max-snippet:0, and any data-nosnippet attribute wrapped around the section you want quoted. A single data-nosnippet on a summary block removes exactly the passage you are trying to get cited.

Also confirm your site is included in Search generative AI features in Search Console, which Google names as a separate eligibility requirement alongside the technical ones.

Step 2. Check nothing is blocking the crawl

Review robots.txt plus any CDN or WAF rules for blocks on Googlebot. Note that Google-Extended governs whether your content is used for Gemini model training and grounding and has no effect on Google Search, including AI Overviews and AI Mode. Blocking it will not remove you from AI Overviews, and allowing it will not add you.

Step 3. Match the heading to the query wording

Use the exact phrasing a person types as an H2 or H3. If the query is "how long does an SEO audit take", the heading is that question, not "Audit timelines". The heading is the retrieval anchor.

Step 4. Write the answer block to fit the space available

Put a complete answer in the first 40 to 70 words beneath that heading. Pew found the typical AI Overview in its sample ran to 67 words, with the shortest at 7 and the longest at 369. Your passage is competing to be part of a paragraph that size, so it has to resolve inside it.

Four rules for the block:

• Open with the answer, not a windup, and never with "it depends".
• Name the subject in the first sentence. A passage beginning with "it" or "this approach" loses its subject when quoted alone.
• Keep sentences under 25 words.
• Put any number, date or source inside the block rather than in the paragraph after it.

Step 5. Re-check after four weeks, then leave it alone

Recrawling and re-synthesis are not instant, and AI Overview composition varies between identical searches. Give it four weeks before judging, then re-run the SERP features filter to see whether the keyword still triggers the feature at all. Do not iterate weekly on a surface this volatile.

What is an AI Overview citation actually worth?

Less than the pre-AI baseline, more than not being cited. Both halves matter.

On the cost side: Ahrefs found the presence of an AI Overview correlated with a 34.5% lower position-one CTR in its April 2025 study of 300,000 keywords, and 58% when it re-ran the same methodology on December 2025 data. Pew tracked real browsing and found users clicked a result link on 8% of searches carrying an AI summary, against 15% without one, with clicks on the summary's own cited sources at 1%.

On the upside: Seer Interactive, analysing 3,119 informational queries across 42 organisations, reported that brands cited inside an AI Overview earned 35% more organic clicks and 91% more paid clicks than uncited brands.

Read together: the AI Overview takes clicks from everyone on the page. Being cited recovers a share of what you lost. It does not restore the pre-AI Overview number, and any vendor implying it does is misreading the studies.

Worked example: sizing the work before you fund it

Illustrative inputs, chosen round rather than flattering. Substitute your own.

Inputs

• Tracked keywords: 400
• Split after the SERP features filter: 104 informational keywords currently trigger an AI Overview; 7 commercial ones do; 289 do not
• In-scope keyword set: 111 of 400, or 28%
• Of those 111, you rank positions 1 to 5 on 34
• Retrofit cost: 34 pages at 40 minutes each, so 23 hours
• Blended internal cost at $60/hour: $1,360 one-off
• Combined current organic clicks from those 34 keywords: 900/month

Arithmetic

• Assume you win citation on half the 34, so 17 keywords, delivering roughly 450 of those 900 clicks
• Apply Seer's cited-versus-uncited gap of 35% to that half: 450 × 35% = +158 clicks/month
• At a 1.0% visitor-to-enquiry rate: +1.6 enquiries/month
• At an illustrative $2,000 gross margin per closed enquiry and a 5% close rate: 1.6 × 5% × $2,000 = $160/month
• Payback on $1,360: 8.5 months

Three assumptions break this: a lower win rate than half, a smaller in-scope keyword set, or a sales cycle longer than the payback window. Run it with your real numbers before approving anything. The point of the model is not the answer, it is that the answer is knowable in advance.

Should you fund AI Overview work at all?

Not always. There are three defensible answers to how to rank in AI Overviews, and one of them is to spend the money elsewhere. Pick on evidence, not on fear of missing out.

That last cell is the one most agency content leaves out. For a meaningful share of B2B companies, AI Overview optimization in 2026 is a genuinely poor use of money, and the keyword audit in step 3 tells you within an afternoon whether you are one of them.

What most teams get wrong about AI Overviews

Adding schema and expecting citation. Google states that structured data is not required for generative AI search and that no special schema.org markup exists for it. Structured data is still worth implementing for rich result eligibility. It is not an AI Overview lever.

Publishing an llms.txt file. Google states its Search systems, including generative AI features, do not use these files, and that publishing one neither helps nor harms Google rankings. It is not a Google tactic. Whether it matters elsewhere is covered in our guide to getting cited by ChatGPT.

Chopping pages into fragments. Google says there is no requirement to break content into small pieces for AI to understand it, and that its systems handle pages covering several topics. Writing self-contained answer blocks is good practice because it is clear writing. Restructuring a whole site into fragments is not.

Changing the date without changing the content. Freshness matters when the underlying facts change. Updating a byline date on a page whose substance has not moved is not a freshness signal, and Google has said as much.

Judging the feature weekly. AI Overview composition varies between identical searches and shifts with product updates. A citation that disappears one week may return the next with no change on your side. Monthly is the shortest sensible review cycle.

Treating a citation as a traffic win. Pew found clicks on the sources cited inside an AI summary occurred on 1% of visits. If you booked AI Overview work on a session-growth forecast, the programme will be judged against a number it was never going to move.

How Euracle approaches AI Overview visibility

The order of operations is the differentiator, not the tactics. Most AI Overview projects begin with formatting. Euracle's begins with subtraction.

The Eureka Method, Euracle's discovery sprint, runs four phases against this problem. Discover applies the keyword split in step 3 above and reports how much of the client's keyword set is actually in scope, which is frequently under a third.

Design decides between the three strategies in the comparison table, in writing, with the arithmetic attached.

Deploy ships answer blocks only on pages already ranking in the top five.

Scale re-measures monthly and reallocates.

The stack is unglamorous on purpose: Ahrefs and SEMrush for the SERP features split and citation tracking, Search Console and GA4 for first-party performance, and n8n with the Claude API to re-run the keyword classification on a schedule so the in-scope list stays current without an analyst repeating it by hand.

Two structural commitments come from how Euracle is built rather than from this discipline. Senior practitioners only: the people in the pitch do the work. And one contract across six disciplines, so a finding that the real problem is rankings rather than formatting does not require renegotiating with a second vendor. Across 50+ projects shipped for 100+ B2B teams in 12 countries, Euracle targets under 90 days from kickoff to first measurable result, and has held 98% client retention over 3 years.

If this is the gap, that is Euracle's AEO and GEO service. The keyword-split problem is sharpest for B2B SaaS companies, where the informational content library is large and the commercial keyword set that actually converts is small.

How do you track AI Overview performance?

Use Search Console's Generative AI performance report, which shows how content is performing in Google's generative AI features on Search and Discover. It is the only first-party data Google publishes on this, and Google cautions against third-party tools that promise ranking success or claim access to internal Google metrics, because none has it.

Alongside it, track two things yourself. First, the size of your in-scope keyword set month over month, from the SERP features filter. A shrinking set means the feature is retreating from your category and the budget should follow. Second, clicks on the specific keywords where you hold a citation, compared against the ones where you do not. That comparison is the closest thing to a controlled test available without vendor data.

The classic fundamentals still carry most of the weight here, which is why this sits inside a wider programme rather than beside one. Our B2B SEO strategy guide covers the ranking work that step 1 of every strategy above depends on, and the generative engine optimization guide covers the surfaces beyond Google.

Conclusion

How to rank in AI Overviews comes down to three decisions, not ten tactics. You can now measure how much of your keyword set is genuinely in scope, decide between the three strategies with arithmetic rather than instinct, and apply the five steps to the pages where they will actually pay. The single most useful thing you can do this week is run the SERP features filter and find out how small your in-scope list really is. If it turns out the real constraint is rankings rather than formatting, that is a more valuable finding than any citation you were going to chase. If you would rather have one senior team run the audit and the work together, talk to Euracle about AEO and GEO.

Sources

1. Ahrefs, 76% of AI Overview Citations Pull From the Top 10, Louise Linehan and Xibeijia Guan, 21 July 2025. https://ahrefs.com/blog/search-rankings-ai-citations/
2. Ahrefs, AI Overviews Reduce Clicks by 34.5%, Ryan Law and Xibeijia Guan, 17 April 2025. https://ahrefs.com/blog/ai-overviews-reduce-clicks/
3. Ahrefs, 2026 update to the above, published February 2026 on December 2025 data, reporting 58%. https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/
4. Google Search Central, AI Features and Your Website. https://developers.google.com/search/docs/appearance/ai-features
5. Google Search Central, Optimizing your website for generative AI features on Google Search, last updated 10 July 2026. https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
6. Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results, 22 July 2025. https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/

Written by

Devanshu Takkar

Product Manager