After AI Search: What Really Changes for Service-Business Websites (and What Doesn't)

What We Know, What Vendors Claim, and What Is a Guess

Almost everything written about 'AI and SEO' blends three different kinds of statement. Separating them is the most useful thing we can do.

What Independent Research Shows About Clicks

The most-cited independent study is from Pew Research Center, published 22 July 2025. Pew analysed the browsing activity of 900 US adults who agreed to a tracking app during March 2025, covering 68,879 unique Google searches. About 18% of those searches produced an AI summary.

On visits where an AI summary appeared, users clicked a traditional search result link in 8% of visits, compared with 15% on visits without one. They clicked a link inside the AI summary itself in 1% of visits. Browsing sessions ended after 26% of pages with an AI summary, compared with 16% of pages without one.

Read this carefully. It shows people click less when an AI summary is present, and that they rarely click the cited sources. It does not show how much traffic any particular website lost, it covers Google only, and it reflects search pages as they appeared in spring 2025. AI features have been expanded since, so the share of searches affected is likely different now. We did not find a newer independent study of comparable quality.

What Vendor Studies Claim (Handle With Care)

SEO-tool vendors publish larger numbers. Several 2026 roundups cite one vendor finding a drop of roughly 61% in organic click-through on queries with AI Overviews and another finding a drop of about 58% for the top organic result. The same roundups report that the losses are concentrated in informational queries, that commercial and transactional queries lose less, and that branded searches fare better. One source reports some partial recovery in early 2026.

We cannot verify these from primary datasets, the methods and time windows differ, and the vendors sell tools whose value rises with alarm. Use them only as a sign that the effect is real and uneven. The pattern, that informational 'how to' queries are the most exposed, is plausible and consistent with Pew's direction.

Importantly, we found no credible 2026 study isolating the effect on local searches such as 'emergency plumber near me', which typically trigger the map pack rather than an AI summary. If someone sells you a number for that, ask for the dataset.

What Google Tells Site Owners

Google's documentation, 'AI features and your website', is unusually direct. It says a page must be indexed and eligible to be shown in Search with a snippet to be a supporting link in AI Overviews or AI Mode, and that there are 'no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.' It also states that you do not need to create machine-readable files, AI text files or special markup.

For control, Google points to the usual tools: robots.txt for Googlebot access, and nosnippet, data-nosnippet, max-snippet or noindex to limit what is shown. It says traffic from these features is counted in Search Console's Performance report under the Web search type.

The practical reading: the ranking and quality fundamentals still decide whether you are a source, and there is no separate 'AI SEO' trick that Google endorses. Google can change its position, but this is the official one today.

• Keep your important pages indexable and snippet-eligible. Do not noindex or nosnippet pages you want cited.
• Check robots.txt does not block Googlebot from pages you want found.
• Watch the Performance report in Search Console for changes in impressions versus clicks. A rising gap is the visible signature of AI summaries.

What About Other AI Crawlers and llms.txt?

Beyond Google, AI companies run crawlers that read websites to train models or to answer questions. Cloudflare's Radar data on crawl-to-referral ratios, how many pages an AI company's crawlers fetch for each visitor they send back, shows a large imbalance, especially for training-focused crawlers. The exact ratios vary enormously between reports; one Search Engine Journal analysis counted at least seven different figures attributed to Cloudflare for a single company. The consistent finding is the direction: training crawlers take far more than they return, and user-triggered and search crawlers return more.

The decision to allow or block these crawlers is a business choice. Allowing them can make your content available to AI answers; blocking them withholds content from training but may reduce your visibility in some AI tools. For most local service businesses, the content in question is public marketing material you want found, so a blanket block usually costs more than it protects. Decide deliberately, crawler by crawler, using robots.txt.

On llms.txt, a proposed file meant to guide AI systems: Google's John Mueller has said Google does not use it, said in 2025 that no AI system currently uses it, and in mid-2026 reportedly repeated that he knows of no confirmed use beyond SEO tools. We maintain one on our own site because it is cheap, but we do not claim it helps and neither should you. If a vendor is selling llms.txt as an AI-visibility strategy, that is a red flag.

Why Local Service Businesses Are in a Different Position

Most of the alarm is about content publishers: sites that earn traffic by answering general questions. A contractor's website is different. Its value is not page views; it is phone calls and bookings from people with an immediate need.

Three things follow, with the caveat that these are our reasoning and not measured results.

• Transactional and local-intent searches ('plumber near me', 'roof repair Denver') are the kind that lead to a call, not a summary read. Vendors report these lose less than informational queries.
• The customer still has to choose who to call. AI answers may shortlist businesses, but they draw on the same signals that already matter: your profile, reviews, reputation and website. Whitespark's 2026 survey shows how heavily Maps ranking leans on category, proximity, hours and reviews.
• Informational blog traffic is the exposed part. If a large share of your site traffic comes from general 'how to' articles that don't lead to calls, expect it to be more volatile. Judge your content by calls and bookings, not visits.

What to Do Now: A Practical Plan

These steps are sensible with or without AI, which is the point.

Three Scenarios for the Next Few Years (Our Forecast, Not Evidence)

Nobody knows how this settles. These are scenarios to plan against, not predictions.

• Scenario A, gradual shift: AI summaries expand but local, high-intent searches still lead to maps and calls. Informational traffic erodes; service businesses with strong profiles and reviews are largely unaffected. This is the outcome the current evidence best supports.
• Scenario B, assistants shortlist providers: AI assistants increasingly recommend and even book local businesses directly. Reviews, structured and accurate business data, and the ability to answer or book instantly become decisive. Businesses that cannot be reached by phone, text or an online booking link lose out.
• Scenario C, fragmentation: customers spread across Google, assistants, social video and marketplaces. Brand strength and presence everywhere matter more than any single channel's ranking.
• A move that holds in all three: be reputable, accurately listed, fast and easy to book. That is why our plan above has no AI-specific tricks.

How We Are Approaching This Ourselves

We are a marketing and automation company with a commercial interest in the answer, so weigh this accordingly. We publish original, sourced guides on this site, keep our own Google profile and listings accurate, and treat the phone and booking flow as part of the website. We will update this article as better evidence appears and note the date when we do.

Sources

Pew Research Center, 'Google users are less likely to click on links when an AI summary appears in the results', 22 July 2025 (pewresearch.org).

Google Search Central, 'AI features and your website' (developers.google.com/search/docs/appearance/ai-features).

Cloudflare Radar crawl-to-referral data and Search Engine Journal analysis of how the figures are reported; Search Engine Journal reporting of John Mueller's statements on llms.txt (2025 and 2026).

Vendor click-through figures (Seer Interactive, Ahrefs and others) as cited by 2026 industry roundups; not independently verified.

Whitespark, Local Search Ranking Factors 2026.