AI Overviews in Google Search: What Changes for Small Sites

AI Overviews put a generated summary with links above the classic results. So a small site now fights for a citation inside that summary, not just for a ranking position. Google announced the feature at Google I/O on May 14, 2024, with a rollout to all users in the US and more countries to follow. Below I keep three things apart: what Google itself has said, what nobody knows yet about traffic, and which controls a publisher really has.

What Are Google AI Overviews and What Did Google Announce?

Google AI Overviews are AI-generated summaries of a topic, shown in Search together with links to learn more. Sounds familiar? It should. Until now the feature ran as an opt-in experiment in Search Labs under the name Search Generative Experience. For US users it is leaving that stage, so people will see it without signing up for anything.

In its announcement of generative AI in Search, the company says the summaries run on a Gemini model customized for Search. It also reports that people used them billions of times during the Labs experiment, and that they use Search more and are more satisfied with their results. Fine. But these are Google’s own statements, not independent measurements. And the post contains no data about clicks to publishers. None.

What Changes on the Results Page for a Small Technical Blog?

For some queries the summary takes the top of the page, which pushes the classic blue links lower. Your page can then play one of two roles: it gets cited as a link inside the overview, or it shows up only in the regular listings underneath. Both can happen to the same URL on different searches (yes, really).

Google describes the use case as wanting “a quick answer” without piecing information together. Short factual questions fit that description best. An in-depth, hands-on guide? Different story. A reader who has to configure a server or debug an error still needs the full page. How all this plays out for click-through rate is not known yet, and it will differ from query to query.

How to Measure the Effect on Your Click-Through Rate

The only evidence I would trust is your own Search Console data, compared before and after mid-May 2024. Other sites’ screenshots tell you little about your niche. You don’t need anything fancy, a simple routine does the job:

1. Note the rollout date in your analytics annotations or a plain text file.
2. Filter the Performance report to the United States, since that is where the feature is live.
3. Compare impressions, clicks and click-through rate per query for equal periods on both sides of the date.
4. Group queries into quick-answer and in-depth types.
5. Recheck after a few weeks, once more data has accumulated.

Watch for queries where impressions hold steady but clicks drop. Those are your candidates for an overview that answers the question right on the results page. And don’t draw conclusions from a few days of numbers. Check the same weeks of last year too, so a seasonal swing doesn’t get mistaken for an effect of the new feature.

What Can You Control? nosnippet, max-snippet and data-nosnippet

Publishers control how much page text Google may show through the existing snippet controls. Nothing new here, honestly. The Google documentation on snippets lists three of them:

• nosnippet - a meta tag that prevents Google from displaying a snippet for the page in search results.
• max-snippet:[number] - a meta tag that specifies the maximum length of your snippets.
• data-nosnippet - an HTML attribute that keeps selected parts of a page out of a snippet.

There is a catch, though. Restricting snippets also removes the normal description under your regular result, and that can cost clicks. So apply these settings per page or per fragment, not site-wide. You set them in your own CMS or page templates, usually through an SEO plugin or the head section of a theme.

One thing people mix up all the time: snippet controls are not crawler rules. Robots.txt decides which bots may fetch your URLs, while these tags decide how much fetched text may be displayed. Our guide to blocking AI crawlers covers the first topic and should not be treated as a switch for the summaries in Search.

Which Content Still Earns the Click When AI Answers in Google Cover the Basics?

Pages that offer something a short summary cannot hold. That’s what keeps giving readers a reason to visit. Full configurations, working code, step-by-step troubleshooting, first-hand detail - all of that belongs here. A definition can be squeezed into two sentences. A tested NGINX config file cannot.

In practice: answer the question in the opening sentences of each section so the page is easy to quote, then go deeper than any overview can. Write an accurate meta description as well, because Google’s snippet page notes that high-quality descriptions can improve both the quality and quantity of search traffic. The same care goes for titles that earn clicks - the regular listing remains your main entry point, after all. Back up claims with primary references too, and treat linking out to sources as part of the page’s value, not as a leak.

What to Do This Month and What to Leave Alone

For a small site the sensible response in May 2024 is, in my view, to measure first and change little. Three priorities are plenty for now:

• Record a baseline of impressions, clicks and CTR for your top queries.
• Review those queries by hand in US results where you can, and note which ones show a summary.
• Improve the pages that answer quick questions thinly, adding the depth a summary lacks.

And what to skip? Some reactions do more harm than good. A blanket nosnippet across the whole site strips every listing of its description. Rewriting everything for a feature that is days old burns effort you cannot yet aim. Acting on anecdotes from other publishers ignores a simple fact: their queries are not yours.

AI Overviews change where your result sits and add a second way to be seen. Their effect on a given blog? Still an open question. Keep a clean baseline, use snippet controls narrowly, and put your time into pages worth opening. That stance holds whatever your data shows in the coming weeks.

FAQ