Key takeaways
- AI Overviews show up mostly on informational, question-style queries, which is exactly where a full answer on your page gives users no reason to click.
- You do not need to rank #1 to get cited, so you can optimize for the citation separately from the click.
- The trick is to give the extractable answer in two or three sentences and keep the useful stuff (data, tools, templates, next steps) behind the click.
- Google is citing its own properties more, especially YouTube, so video and Google-owned surfaces deserve a place in the plan.
- Measure citations, clicks, and conversions per page. Some pages are worth sacrificing, others are not.
The problem: the better your answer, the fewer clicks you get
Here is the awkward part of this whole topic. Everything the standard advice tells you to do (lead with a direct answer, write a TL;DR, add an FAQ) makes your content easier for Google to lift and summarize. Do it badly and you have written the AI Overview for Google, then watched your sessions drop.
The numbers are not comforting. Pew Research tracked 68,879 real Google searches from 900 U.S. adults and found that 8% of users who saw an AI Overview clicked a traditional result, versus 15% when no AI summary appeared. People treat the summary as the destination.
So the goal is not simply "rank in AI Overviews." It is to get cited for the right queries, with the right amount of information given away, and to make the click worth taking for the reader who wants more than a summary.
Which informational queries trigger AI Overviews
Ahrefs analyzed 146 million SERPs and found AI Overviews on about 21% of all keywords, but far more often on certain query shapes:
- Question queries trigger them 57.9% of the time, versus 15.5% for non-questions.
- Queries with 7 or more words trigger them 46.4% of the time, up from 9.5% for one-word queries.
- "Why" queries lead at 59.8%, followed by yes/no queries (57.4%) and definition queries (47.3%).
- Nearly all AI Overview keywords are informational. Commercial and transactional queries are a tiny slice of the total.
A caveat: the claim that 99.9% of informational keywords trigger an Overview, which circulates in some summaries of this research, is a misreading. The Ahrefs figure is that 99.9% of the keywords that do have an Overview are informational. Those are very different statements, and the second is the one the data supports.
The practical read: definition and "why" queries are the ones most likely to be fully answered on the results page. Those are also the queries where a click was least likely to begin with.
Step 1: Sort your informational pages into three buckets
Before you touch a single page, decide what each one is for. I use three buckets.
| Bucket | Typical query | What the page is for | Approach |
|---|---|---|---|
| Citation pages | "What is X", "Is X safe" | Brand exposure, not clicks | Answer fully and briefly, make the brand name visible, accept low CTR |
| Bridge pages | "How to do X", "Why does X happen" | Give the answer, sell the depth | Short answer up top, tools, data, and templates below |
| Money pages | "X vs Y", "best X for Z" | Conversion | Rarely a pure AI Overview target, protect these |
Most sites treat every informational page as a Citation page by accident. That is the traffic leak. A "how to" query where the answer is a 12-step process is a Bridge page: Google can summarize the idea, but the user needs your checklist, calculator, or screenshots to actually do it.
Step 2: Write the answer for the machine, the value for the human
The structure that earns citations is well documented. Seobility's checklist puts it plainly: lead with a clear, direct answer, use a logical heading structure, and add TL;DRs and FAQs. Google's own Search Central guidance on succeeding in AI search says much the same thing about helpful, people-first content.
What I would add is a rule about how much to put in that opening block.
The two-sentence rule
Give the extractable answer in two or three sentences. State the conclusion, not the method. For example:
- Extractable: "Query fan-out is when an AI system splits one question into several searches and merges the results."
- Not extractable (and worth the click): the list of fan-out queries we saw for 40 real prompts, the spreadsheet to reproduce it, and how it changed after the last model update.
The first line is what Google quotes. The second is why someone opens your page.
Put the gated value in a recognizable place
Tell the reader, in the page itself, what is further down. A line such as "Below: the full 14-point checklist and a downloadable template" is both good UX and a nudge for the user who sees a cited snippet and wonders whether there is more.
What to hold back
- Original data tables and benchmark numbers (the summary can mention the headline, the table lives on your page)
- Step-by-step screenshots
- Calculators, templates, and interactive tools
- Opinionated recommendations with reasoning
- Comparisons with your own testing
These are also the things an AI summary handles worst, so keeping them on your page costs you nothing in citation chances.
Step 3: Target the long-tail questions Google cannot fully answer
Long, specific questions trigger Overviews often, but they are also the questions where the summary is most likely to be thin or hedged. A page that clearly answers "how does retrieval-augmented generation affect AI Overview source selection" is more likely to be cited than a vague overview of AI search, as SEOcrawl puts it, and it keeps more clicks because the reader is a specialist with a real problem.
A workable research loop:
- Pull question keywords in your niche, filter for 7+ words, informational intent, and check which already trigger an Overview. Ahrefs describes exactly this workflow in Keywords Explorer.
- Add People Also Ask and autocomplete questions.
- Mark each as Citation, Bridge, or Money.
- Only write Citation pages when the brand exposure is worth it.
For question mining on a smaller budget, AlsoAsked and AnswerThePublic show the real follow-up questions people ask.
Step 4: Build the follow-up, not just the answer
This is the part most guides skip. An AI Overview ends with a question mark in the user's head. What comes next? If you can predict that, you can write the page that gets the second search.
Take "why is my website traffic dropping." The Overview lists five causes. The reader now wonders which of them applies to them. A page that offers a diagnostic flow, a Search Console walkthrough, and a checklist by symptom is not competing with the Overview. It is the next step after it.
Practical ways to do this:
- End every Bridge page with the specific question the reader will have next, and link to the answer.
- Build topic clusters with real internal links. Seobility notes that clustering makes it easier for Google to recognize your site as a comprehensive source, which helps with query fan-out, where Google runs several related searches behind one question.
- Make the follow-up page the one with the tool, template, or data.
Step 5: Do not only rely on text formats
Promptwatch Data tracked what Google AI Overviews cited in July 2026, and the content mix is more varied than most plans assume:
| Content type | Share of AI Overview citations (July 2026) |
|---|---|
| Listicle | 18.0% |
| Product page | 16.3% |
| How-to | 15.1% |
| News article | 13.5% |
| Video | 5.9% |
| Social post | 5.1% |
| Landing page | 4.7% |
| Comparison | 3.6% |
Source: Promptwatch's AI Overviews citation types report for July 2026. Listicles averaged about 26% in Q1 and are around 18% now, while product pages roughly doubled from about 9% in January. Video reached about 6.3% by the end of July, up from about 2.7% in January.
Two takeaways for traffic protection. First, how-tos are 15% of citations, and they are the format where a summary is least satisfying, so they are the best Bridge pages. Second, the shift toward product pages and video means informational content is not the only way into the Overview.
Google is citing Google
In Promptwatch's June 2026 AI Overview citation share data, YouTube held 4.16% of citations, Reddit 2.72%, and google.com 2.30%, all year-to-date highs. google.com went from 0.37% in January to 2.30% in June. YouTube and google.com together are about 6.5% of all sources.
That has a direct consequence for your plan. When Google answers with its own properties, a blog post is competing against a surface the click already lives on. Promptwatch's advice from that report is to treat Google surfaces as assets (a YouTube channel, Business Profile, help content) rather than fight for the same citation.
Social is not equal across engines
From Sep 28 to Oct 4, 2026, social platforms were 12.2% of AI Overview citations, compared with 0.46% for ChatGPT. Inside AI Overviews, YouTube was 4.28%, Facebook 2.38% (now ahead of Reddit at 2.15%), Instagram 1.61%, and LinkedIn 1.55%. If your audience searches on Google, a YouTube video that answers the question and points to your page is a legitimate second route into the same Overview. I would not count on Reddit to do the same job in ChatGPT anymore.
A video also gives you something a text summary cannot: a thumbnail, a creator face, and a reason to click through. Tools like Pictory or Descript-style editors make repurposing a how-to article into a short video cheap.
Step 6: Technical hygiene, kept short
None of this works if Google cannot read the page. The checklist is the usual one:
- Pages are indexable and in the sitemap, with no stray noindex tags.
- Content is in the HTML, not rendered late behind JavaScript. If your site is a JS app, a prerendering service helps.
- Structured data matches what is visible on the page. Do not add FAQ markup for questions you do not answer.
- Core Web Vitals are not embarrassing.
- Author, date, and update information is visible.
Google Search Console is where you check indexing, and it also gives you the click data you will need in the next section.
For crawling your own site and catching the mistakes above, Screaming Frog is the standard.
Step 7: Measure what a citation is actually worth
This is where "without giving away your traffic" becomes a number rather than a feeling. For each informational page you care about, track:
- Whether it is cited in an AI Overview, and for which queries
- Impressions, clicks, and CTR in Search Console for those queries
- Conversions or assisted conversions from the page
- Branded search volume over time, since exposure in the Overview often shows up there first
Semrush can show whether your domain appears in AI Overviews through the Organic Rankings report and its SERP Features filter.
For a dedicated view of which of your pages get cited, in which AI engines, and what traffic that brings, a visibility platform is more practical than checking by hand. Promptwatch tracks Google AI Overviews alongside ChatGPT, Perplexity, and others, joins citations with Search Console data per page, and flags pages that rank but never get cited.
If you are comparing options, the directory at bestgeosoftware.com lists GEO and AI visibility tools side by side.
A simple decision matrix helps:
| Cited? | Clicks | Conversions | What to do |
|---|---|---|---|
| Yes | Falling | Flat | Fine, you are buying exposure. Check branded search |
| Yes | Falling | Falling | Move the valuable part of the answer below the fold, strengthen the bridge |
| No | Falling | Falling | The Overview is using competitors. Revise the opening answer and structure |
| Yes | Stable | Rising | Leave it alone |
Step 8: Refresh deliberately
A lot of AI Overview citations come from news and how-to content, and recency matters for both. Build a refresh rhythm, not a rewrite habit. Update the numbers, the date, and the one section that has gone stale, and leave the rest.
For spotting which pages have changed or slipped, a change-detection tool is useful for watching competitor pages as well as your own.
Common mistakes
- Answering the query completely in the first paragraph for every page, including ones that should drive a click.
- Treating FAQ schema as a magic wand. Markup that does not match visible content does not help.
- Writing only listicles because they are the biggest cited type. Their share is falling.
- Ignoring video and Google surfaces while Google cites itself more.
- Judging success by sessions alone. A cited page can be doing its job with fewer clicks.
- Optimizing a page for citation when it is actually a money page.
A simple 30-day plan
- Week 1: Export your informational pages from Search Console, label them Citation, Bridge, or Money.
- Week 2: Rewrite the openings of your top 10 Bridge pages with a two-sentence answer and a clear pointer to what is below.
- Week 3: Add or build the follow-up page, tool, or data table for each of those pages. Record one short video for the best of them.
- Week 4: Start tracking citations and conversions per page. Revisit the decision matrix above.
If you would rather hand this to specialists, an agency that does technical SEO and GEO together can run the audit and measurement for you. 1001 SEO Media does that kind of work, though the framework above is something a small in-house team can run itself.
Final thought
You cannot stop Google from summarizing the easy part of your content. You can decide what the easy part is. Keep it short, make it quotable, and put the things worth clicking for (data, tools, judgment) where a summary cannot reach.