What have the authority sites measured recently?

On 2026-07-02, Semrush published “AI Overviews are expanding across commercial intent search.” The core message: AI summary coverage has widened from informational questions to commercial-intent queries. Commercial intent means the terms a user searches when they are already comparing options and getting ready to decide.

For export sites, this matters more than the previous wave. Having informational terms covered is tolerable, because they never brought orders directly; but when “who is best, what price level, how do I choose” get intercepted by a summary, the stretch of road closest to the sale is partly blocked as well. Ahrefs’ French-market test on 2026-08-27 gives a sense of the scale: a 23.1% CTR decline when AIO appears. That is not zero, but it is enough to change how you allocate budget.

At the other end, on 2026-10-01 Ahrefs measured AI Overviews appearing for 83% of branded searches. Put the three findings together: summaries have moved from informational terms to commercial terms, and from non-brand terms to brand terms — coverage is still spreading. I keep tracking this on the AI Overview click decline thread.

What does it mean when commercial terms get covered?

First, separate two kinds of coverage. In the first, the summary answers the user’s question outright and the user does not need to click any site. This kind hurts most, and it usually happens on terms where the answer is standardized and there is no room for trade-offs. In the second, the summary points in a general direction but the user still has to click through for specific conditions and a quote. This kind still leaves an opening, because no buyer places an order off a paragraph of summary.

Among commercial terms, the first kind is usually a loosely bounded question like “what does XX cost” or “which is better, XX or XX”; the second is “given my specification, my volume, and this lead-time requirement, what is the economical way to do it.” The more specific the buying question, the harder it is for a summary to finish answering it alone. The job for an export site is to move its content toward the second kind.

One more point that is easily overlooked: the sources inside a summary are usually just a handful. If your page has never been summarized, that does not mean you have no chance; it may be that the page structure leaves the model unable to extract a usable sentence. That is a technical problem: start with Search Console diagnostics to look at crawling and impressions, then decide whether to change the content or the structure.

There is a balance to strike here. Writing specifically does not mean writing long and convoluted. Summaries favor short sentences and clear conclusions; too long and too scattered, and no key point can be extracted. My approach is to open each page with two or three sentences that answer directly, with the conditions and ranges written in, then use subheadings to expand on the reasons and the cases. That way you sacrifice neither depth nor the summary’s ability to read you. Another common mistake is packing everything into one giant table, so that neither the reader nor the model can tell what matters; splitting it into two or three smaller tables is clearer. When I revise a draft I first count how many “direct answer” paragraphs the page has. If a page has only one, it is too thin for a summary to work with.

Our Chinese-language test: which buying keywords still have gaps?

I ran an AIO coverage test on 500 Chinese-language queries: an 80.9% trigger rate overall, 84.1% in the US region and 77.8% in the Hong Kong region. The difference by keyword type was the most useful finding here: 96.7% for tutorial terms and 62.9% for technical terms. The full record is in our measured Chinese AI Overview coverage.

That 34-point gap is where the opening lies. Technical terms are not fully covered because they require conditions, parameters, and experienced judgment that a summary cannot supply in full. For sites in categories like machinery, chemicals, and building materials, this means that writing content as technical judgment rather than general explainer content can still earn clicks.

There is a colder set of numbers to set beside that: this site’s indexation rate is 100%, yet in 28 days it drew only 12 impressions and 0 clicks. Indexation is not the problem; being selected is. If your site also indexes well but earns few impressions, do not rush to build more pages — first check whether the titles and openings of your existing pages answer a real question. You can work through them page by page against title, URL, and paragraph priority.

Which three parts of a covered page should you change?

Give your judgment in the opening paragraph

Do not set up the industry background first. Answer the question in the title in the first paragraph, then write the reasons and conditions. A summary has to judge what the page is about in a very short space; the earlier your conclusion appears, the higher the chance it gets restated correctly.

Replace vague answers with conditional ones

“Price depends on the configuration” says nothing; “the standard specification range is X to Y, and material and surface treatment matter most” is useful. Sentences with ranges and conditions are easy for a summary and easy for a buyer to use when comparing you with another supplier. For how to write them, see the four-part content template.

Add comparisons and trade-offs

The deeper need behind a commercial term is comparison. A real comparison table that spells out where each option fits and where it does not is far more useful than another paragraph of adjectives. Product page titles and descriptions should be written the same way — see product page titles, descriptions, and FAQ structured data.

How do you verify that the changes worked?

Do not look only at total traffic. Record the summary status and your impressions and clicks for the same set of terms before and after the change, with four weeks as the observation window. If the summary for a term is still there but your impressions went from zero to something, you have entered the candidate set; if impressions did not move, the opening paragraph is not direct enough yet.

Look at page type as well. Ahrefs’ research suggests that the impact differs by page type, and I worked through a round of this in breaking down AIO traffic loss by page type. Separating blog posts, product pages, and category pages tells you which type to rescue first.

Finally, keep a record of counterexamples: any page that got worse after the change. If you note only the successes, you will mistake random fluctuation for method. That habit matters more than any tool; read it alongside the traffic-decline case review.

A low-cost action checklist

Three steps; the first round takes an afternoon.

1. List 20 core commercial terms and check each one for a summary: note whether a summary appears, who it mentions, and whether it links to you.
2. Pick the 3 terms closest to a sale and rewrite their openings: move the conclusion up front, add ranges and conditions, and leave the rest of the site alone.
3. Check the same set of terms again four weeks later: compare impressions and clicks, and write down your explanation and your counterexamples.

Once these three steps are done, you will have a realistic tiering of the terms you hold. For a more systematic view of the overall strategy, read the introduction to answer engine optimization and how a small team should start.

The questions owners ask most often

If a buying keyword gets covered by an AI summary, is it hopeless?

No. What gets covered are informational questions. Buying questions that carry operating conditions, specifications, or comparisons still require a click into the source to check. The key is not to write a page that a single sentence can answer.

How do I find out whether my keywords have an AI Overview?

The easiest way is manual sampling: list your main keywords and search each one in your target market, noting whether a summary appears and who it mentions. A workable sample is enough; you do not need to run the whole site.

Does adding FAQ structured data to a page help?

It helps, but it is not a cure-all. Structured data makes a page easier to understand, provided the answer itself is written specifically. Markup alone with empty content will not be adopted either.

Should I write a separate set of content just for AI?

No. If you write clearly for the buyer, AI will naturally understand it. Writing a separate set for machines usually fails on both counts.