What has actually changed

For most of its history, a search engine answered a question by handing you a list of places where the answer might be. The judgement was yours. Increasingly, the engine attempts the answer itself, assembling it from multiple sources and presenting it above the list.

For a business, three consequences follow.

  • Some questions no longer produce a visit. If the answer is complete on the results page, there is nothing to click.
  • Being cited becomes a distinct goal from ranking. A page can rank well and not be used in the generated answer, or be used without being prominent in the list.
  • Being described wrongly becomes a real risk. Previously, incorrect information about you sat on some page a customer might not visit. Now it may be summarised authoritatively with no obvious way to challenge it.

Which searches are most affected

Query intent is useful for analysis, but there is no reliable rule that informational searches always lose visits or local and transactional searches are protected.

  • Information: a summary may answer a simple question, while a detailed explanation may still earn a visit.
  • Research: comparisons may include generated summaries and links; readers may continue to specifications or pricing.
  • Local or transactional: people still need accurate business details, but the presentation and click path can also change.

Classify your own queries and landing pages, then observe impressions, clicks and qualified enquiries. Avoid projecting a universal traffic-loss percentage onto a particular business.

What this means for clicks

Google includes AI-feature traffic within the Web search totals in Search Console. Those totals do not by themselves isolate the effect of a generated answer.

Compare periods carefully: seasonality, changed rankings, query mix, advertising and tracking changes may all affect results. Use page-level clicks, enquiry quality and conversion data together. A drop in clicks does not establish a single cause, and more citations do not establish more sales.

What still produces enquiries

The channels that hold up best are the ones that answer a question a summary cannot.

  • Local search and the map results. Someone who needs a business nearby needs a specific business, with a phone number, reviews and directions. Check whether this channel produces qualified enquiries for your eligible business, and maintain your Google Business Profile.
  • Branded search. When people search your name, they are looking for you specifically. This grows with reputation and offline presence, and its results can also include generated summaries.
  • Pages with information only you have. Your prices, your availability, your capacity, your specific work. Publish them accurately; generated answers may still misrepresent them.
  • Deep, specific content. The more a question requires nuance, judgement or worked detail, the less a short generated answer satisfies it.
  • Direct channels. WhatsApp, email lists, repeat customers and referrals do not depend on a results page at all.

How a business should respond

Not with panic, and not with a rebuild. Five adjustments cover most of it.

  1. Audit your query mix. In Search Console, look at which queries bring clicks and classify them: informational, research, or transactional. Compare changes by group rather than assuming a group is safe or certain to decline.
  2. Shift content towards what only you can say. Fewer general explainers, more specific material grounded in your actual work, pricing and experience.
  3. Make your business information easier to verify. Consistent details everywhere, current prices, accurate structured data. This reduces the risk of being misdescribed.
  4. Strengthen the channels that do not depend on a results page. Reviews, direct messaging, repeat contact, referral.
  5. Check how you are described. Every quarter, ask a handful of realistic questions in a few assistants and record the answers. Record both absence and inaccuracies; neither is evidence of what all users see.

What not to do

Several reactions to this shift are worse than doing nothing.

  • Do not abandon SEO. Maintain useful search work and evaluate its contribution using your own data.
  • Do not buy a guaranteed AI visibility product. Nobody can guarantee inclusion in a generated answer. No platform submission or dashboard guarantees inclusion.
  • Do not mass-produce AI-written content to compete. Volume without substance was already a weak strategy; against systems that summarise, it is a worse one.
  • Do not block every AI crawler reflexively. It is a legitimate choice for a publisher whose business is selling access to content. For a business that wants to be found and recommended, check each crawler's documented role: search retrieval and model training can have different controls.
  • Do not assume it will settle. Reviewing your query mix once a quarter is a reasonable ongoing habit rather than a one-time project.

The underlying pattern

A useful response is to maintain clear commercial pages, answer distinct customer questions and keep source facts current. No content strategy is immune to platform changes.

For the implementation checklist use AI search optimization; for choosing the scope of work use SEO vs GEO. This article concerns measurement and changing search journeys, rather than a separate set of guaranteed ranking tactics.

Sources and further reading

These sources explain platform guidance or engineering principles, not evidence of PresenticAI client results. Guidance checked on 14 October 2026.

Not sure how exposed your search traffic is?

PresenticAI audits your query mix, tells you which of your searches are actually at risk, and focuses effort on the visibility that still produces enquiries.

See our SEO and GEO services

PresenticAI Editorial Team

PresenticAI is an AI and IT solutions company. These guides cover AI automation, agents, chatbots, web systems and digital marketing, including specialist SEO and GEO topics. Examples are illustrative unless explicitly supported by evidence; they are not client case studies.