Table of Contents
TL;DR
- Ahrefs found a 34.5 percent CTR drop at position one across 300,000 informational keywords when an AI Overview is present. The post still ranks. The reader simply reads the summary and closes the tab.
- Pew Research tracked the same gravity from a different angle and found that only 8 percent of AI-summary visits produced a click on a traditional search result, against 15 percent when no AI summary appeared.
- The rewrite question is not whether to chase the click back, but whether your post is being extracted into the summary the reader actually reads.
- A rewrite for AI Overviews tightens the first 50 to 90 words after every H2 into an answer-first paragraph anchored on a number, a year, and a named source. The supporting story stays underneath.
- Ship the rewrite on one post first, measure with impressions plus citation presence plus branded search, and decide on the next post from a real signal rather than a guess.
Ahrefs measured a 34.5 percent CTR reduction on the position-1 result across 300,000 informational keywords when an AI Overview is present, comparing aggregated Google Search Console data from March 2024 against March 2025.
That single number reframes the rewrite question for a small site. The rank line is not the problem. The post still ranks. What changed is that the reader gets their answer above the rank line and never reaches the page.
If you have ever stared at a Search Console chart where impressions hold steady and clicks slope downward on an old post, the gap has a name now. It is the AI Overview reading the answer to your reader, on your behalf, without sending the reader on.
The right question is no longer "should I update the keywords on this old post?" It is "should I rewrite this old post so the AI Overview cites it as the answer it summarises?"
That second question has a different answer for every post, and the rest of this article works through how to decide.
What does an AI Overview actually do to your old post?
An AI Overview is the AI-generated summary Google now places at the top of the results page for many informational queries.
When that summary appears, the click rate on the top organic result falls hard. The 34.5 percent figure from the Ahrefs April 2025 study measured position-1 specifically across 300,000 informational keywords. The post itself still ranks. The reader simply reads the summary and closes the tab.
The mechanism is not a penalty against your page. It is a shift in where the answer surfaces.
Your post might still be the top-ranked source. The AI Overview reads your post, your H2s, your bullet points. It synthesises a three-sentence summary that answers the reader’s question on the search page. The click never fires.
Pew Research tracked the same gravity at the session level. They followed 68,879 unique Google searches from 900 US adults during March 2025. When an AI summary appeared, only 8 percent of visits produced a click on a traditional search result, against 15 percent when no AI summary appeared. Only 1 percent of AI-summary visits produced a click on a source cited inside the summary itself.
These two numbers describe one phenomenon. The reader gets the answer, and the page that wrote the answer rarely gets the visit. The same gravity drives the broader picture of zero-click search on small sites, which is the wider frame this post sits inside.
Which posts on your site are worth rewriting first?
Three signals tell you which posts to touch first.
Informational queries that trigger an AI Overview on the live SERP. You run the post’s primary query in an incognito window today and an AI summary appears at the top. That is the strongest signal. Posts whose queries do not trigger AI Overviews are not in the rewrite shortlist yet.
Posts where Search Console shows steady impressions and falling clicks over the last twelve months. The decoupling tells you the post still ranks but the click economy has shifted away from it.
Posts whose primary query a reader would type into ChatGPT or Perplexity as easily as into Google. Those queries belong to the AI-Overview territory whether or not Google has rolled the feature out for the exact phrasing yet.
A post that fails all three tests can stay as it is. The classical ranking is still doing the work, and the reader still clicks.
A post that passes all three tests is on the shortlist. Two of three is a strong candidate. One of three is a watch-list entry.
What does a rewrite for AI Overviews actually change in the post?
A rewrite for AI Overviews changes the first 50 to 90 words after every H2 into an extractable answer.
A specific number. A year. A named source.
One clean single-sentence claim the chunker can lift out of the page and quote inside its synthesised answer. The supporting story stays underneath, where it earns the click from the small share of readers who want the depth.
The H2 itself changes from a generic label into the real question a reader would type into the search box. "Putting it together" becomes "How do you measure whether the rewrite actually worked?" One is a tutorial-summary header an AI chunker cannot use. The other is the question the AI Overview was already trying to answer.
The closing CTA stays calm and human.
You do not need to retire any of the post’s existing material. Most of what is already there will sit, unchanged, beneath the answer-first paragraph. The rewrite is the first 80 words plus a few headings, not a full rebuild.
How do you write an answer block that survives the extraction?
An answer block survives extraction when its first paragraph reads complete in isolation.
No pronoun pointing at the prior section. No phrase like "as mentioned above." Concrete subject, concrete claim, the supporting number with the year and the source name inside the paragraph itself.
The block reads as a standalone answer when copied alone into a Word document. That is the same test the AI chunker runs when it builds the summary. The chunker extracts a paragraph, drops the context around it, and asks whether the paragraph still resolves its own claim. Most prose written for a human reader fails this test because it leans on the paragraph above it for context.
The fix is mechanical, not creative. You add the entity name, you add the year, you add the source. You rewrite "they found a 35 percent drop" into "Ahrefs found a 34.5 percent drop in April 2025." You move "this means" out of the answer paragraph and into a follow-up sentence. The structural changes are the same ones a featured-snippet rewrite teaches, and that rewrite shape has been the right shape since long before AI Overviews existed.
What if the rewrite drops the post in the ranked list?
The ranked position holds in most cases because the structural changes a rewrite makes are the same changes that earn citations.
Cleaner H2 questions, an answer-first paragraph, an external source citation in a supporting paragraph, a specific stat anchor. None of those moves contradict what classical ranking already rewards.
The risk is uneven, not symmetrical.
A rewrite that adds clarity tends to hold rank, because the engine reads the same signals as a thoughtful reader. The post becomes easier to scan, easier to extract, and easier to trust. A rewrite that thins the post in search of keyword density tends to lose rank, because the engine reads the thinning the same way the reader does.
The mistake to avoid is treating an AI Overview rewrite as a 2018-era SEO rewrite. The old keyword-density playbook makes the post worse for both audiences. The right rewrite is the one that would help a human reader if no AI engine existed, and that rewrite also happens to be what the chunker can extract cleanly.
How do you measure whether a rewrite actually worked?
Three signals together replace the click-only check.
Impressions on the target query in Search Console, read once a month. The presence of your URL inside the live AI Overview on the target query, read directly by running the query yourself once a month from an incognito window. Branded-search volume on a 90-day window, where the brand impression from the AI summary surfaces six to twelve weeks after the rewrite.
If impressions and citation presence rise together, the rewrite is working even when the click line stays flat. The AI summary is doing the work, and your name is inside it.
If impressions rose but citation presence did not, the answer block is not extractable enough yet. The next pass tightens the first 80 words under each H2 further and adds a clearer stat anchor.
If neither signal moved, the query may not trigger an AI Overview at all yet. The post was not on the right shortlist. You move it back to the watch list and pick a different candidate.
The same three signals show up cleanly in Search Console’s AI Mode data once you know which numbers to read.
Other questions worth answering
What share of cited sources inside a Google summary fall outside the classical top ten?
Roughly 38 percent. The cited sources inside a Google summary often fall outside the classical top ten for the same query. The way Google picks pages to quote is not the same as the way Google ranks them. A page can earn a quote on clarity even when its classical position sits well below the top ten.
Does query length predict if a Google summary appears on the results page?
Yes, and the gap is sharp. Pew Research tracked 68,879 Google searches in March 2025. A summary appeared on 8 percent of very short searches. Question searches drew one 60 percent of the time, and searches of ten or more words reached 53 percent.
What signals identify candidate pages when Search Console shows no impression history yet?
Three signals point to a candidate even when Search Console is empty. First, the live SERP for the target query shows a summary panel at the top. Second, the same query reads well in ChatGPT or Perplexity as easily as in Google. Third, the topic falls in a category where summaries trigger heavily, like education or how-to.
How does Google’s conversational search tab handle the same query that triggers a summary panel?
Google routes the query into a fuller back-and-forth experience instead of a single summary card. The conversational tab keeps context across follow-up questions, which the summary panel does not. As of January 2026, tapping ‘Show more’ on the summary panel on mobile transfers the user to that conversational tab with the question already loaded.
Which rewrite should you ship first?
Ship the rewrite on the single post whose primary query already triggers an AI Overview on a real Google search today.
Open the post. Rewrite the first 80 words after every H2 into an answer-first paragraph with a number, a year, and a named source. Tighten the H2s into the real questions a reader would type. Re-publish, then re-run the query in an incognito window every two weeks.
One post is enough to learn the pattern. You will see whether the rewrite shape feels heavy, whether your H2s drift back into tutorial-summary labels, and whether the answer-first paragraph reads natural to you. Pattern-learning on one post costs an afternoon. Pattern-learning on twenty posts at once costs a quarter and produces uneven quality.
If the first rewrite earns a citation in the live AI Overview within six to eight weeks, the second rewrite is the next post on the same shortlist. If the first rewrite did not earn a citation, the post may not have been the right candidate, or the answer block may need a second pass.
If you read this and the rewrite shortlist on your site still feels foggy, that is normal. Most small-site owners are looking at the same flat click line and the same handful of old posts and trying to guess the order. You can contact me here and we can open Search Console together for one or two posts, run the AI-Overview presence check live, and pick the single post worth rewriting first.
There is no charge for the call, and there is no pitch at the end. I will tell you which posts already have a real AI-Overview signal you can act on, which ones are watch-list candidates, and the single post whose answer-first rewrite is most likely to earn a citation in the next eight weeks.