A business can rank prominently in Google and still become less visible in the way prospective customers make decisions.
That sounds contradictory. For years, search engine optimisation has centred on improving rankings, attracting clicks and converting website visitors.
But the relationship between being found and being visited is changing.
AI-generated search experiences increasingly answer questions directly, compare alternatives and provide information before a user reaches a company's website.
For businesses, the commercial implications extend beyond organic traffic.
The important question is becoming:
Is the business being represented accurately when customers research their options, even when they never visit its website?
The evidence suggests a measurable change in search behaviour
In July 2025, the Pew Research Center published an analysis of Google search behaviour using browsing data from 900 US adults collected during March 2025.
Its findings revealed a significant difference between searches containing AI-generated summaries and those displaying traditional results.
When an AI summary appeared, users clicked a traditional search result in 8% of visits. When no AI summary appeared, that proportion was 15%.
Users clicked a source link within an AI summary in just 1% of visits containing one.
Pew also found that 26% of visits to search pages with AI summaries ended the browsing session, compared with 16% on pages without an AI summary.
The findings do not prove that AI summaries caused every reduction in clicking. Search queries that trigger summaries may differ systematically from those that do not, and the study involved US participants rather than UK businesses.
Nevertheless, the research provides an important observation.
Search can increasingly satisfy information needs without generating a website visit.
For companies evaluating their online visibility, organic traffic therefore becomes a less complete measure of their presence in the customer's decision-making process.
Source: Pew Research Center, July 2025
Visibility and traffic are no longer interchangeable
Consider two specialist accountancy firms.
Both provide similar services and operate in the same region.
The first has a technically sound website, established search rankings and a consistent stream of organic traffic.
The second also appears in industry publications, provides detailed explanations of its specialist services and has independently verifiable information about its expertise.
When someone searches for a straightforward service definition, both businesses may compete for conventional organic clicks.
When that person asks an AI system to identify suitable providers, compare their specialisms or explain which firm may be appropriate for a particular requirement, other factors become relevant.
The system must identify the providers, interpret their services and draw on information available to it.
That does not guarantee the second firm will be recommended. Different AI search products use different retrieval systems, and source selection can vary considerably.
But it illustrates why visibility cannot be reduced to a single ranking position.
A company may need to be both discoverable and sufficiently well described for search systems to understand what it does.
Longer, more detailed questions matter
The Pew study also examined the types of searches associated with AI summaries.
Only 8% of searches containing one or two words generated an AI summary.
For searches containing ten words or more, the proportion rose to 53%.
Queries beginning with question words such as who, what, when and why generated an AI summary in 60% of cases.
These figures describe the search environment examined during March 2025, rather than a permanent rule governing Google's products.
Still, the distinction is commercially important.
Short searches often identify a category.
Longer searches can express a more developed purchasing problem.
Compare:
"accountancy firm London"
with:
"Which accountancy firms specialise in supporting UK businesses expanding into European markets?"
The second question reveals a specific requirement, a market context and potentially a more advanced stage of consideration.
For specialist B2B organisations, those detailed questions may be disproportionately valuable.
A business does not necessarily need to appear for every general search. It needs to be understood when a prospective buyer describes the exact requirement it can fulfil.
The answer is not to abandon conventional SEO
There is a temptation to treat AI search as an entirely separate discipline requiring a new collection of technical tactics.
Google's own guidance advises against that assumption.
Its documentation on optimising for generative AI search says that established SEO principles remain relevant because Google's AI search experiences use its core search and retrieval systems.
Google also emphasises useful, original content and technical accessibility.
There is no special requirement to create a separate AI-only website, add a dedicated AI schema type or rewrite every page into short question-and-answer fragments.
Indeed, Google explicitly cautions against prioritising superficial generative-engine optimisation tactics over useful content.
Source: Google Search Central, Optimising for Generative AI Features
For businesses, the practical foundation remains familiar:
A website must be accessible, its services must be described accurately and its content must address questions that potential customers genuinely ask.
What changes is the need to examine how that information is interpreted and represented across different discovery environments.
A practical framework for measuring AI search visibility
Traditional search reporting typically measures rankings, impressions, clicks and conversions.
AI visibility introduces additional questions.
A useful assessment can be organised around four dimensions.
1. Discoverability
Can the business be identified when users search for its category, specialism, geography or the problems it solves?
This includes conventional search performance and visibility within AI-generated responses.
2. Accuracy
When a system describes the company, is that description correct?
Does it understand the services offered, relevant limitations, location and intended customer?
Being mentioned inaccurately can be commercially unhelpful.
3. Supporting evidence
What sources support the description of the business?
Relevant evidence might include its own website, detailed service documentation, professional profiles, credible third-party coverage and other independently verifiable information.
These are potential sources of context, not guaranteed mechanisms for appearing in AI answers.
4. Commercial relevance
Does the business appear for questions that its intended customers actually ask?
A mention for a broad informational query may be less valuable than accurate inclusion in a specialised purchasing discussion.
These measures should be tracked across a consistent set of queries, with repeated observations over time.
Because AI-generated outputs can vary between sessions, one-off screenshots should not be treated as definitive evidence of visibility.
A worked example: 40 commercially relevant questions
Suppose a specialist consultancy identifies 40 questions that prospective customers regularly ask.
They include questions about service suitability, alternative providers, geographic coverage, pricing models and technical requirements.
The consultancy then tests those same questions across selected AI search environments at regular intervals.
For each response, it records whether the business was mentioned, accurately described and supported by relevant sources.
Imagine the business appears in eight of the 40 responses during the initial assessment.
Its observed inclusion rate is 20%.
Following improvements to its service content and supporting information, it appears in 14 of the same 40 questions during a later assessment.
Its observed inclusion rate is now 35%.
That is an improvement of 15 percentage points within the measured sample.
It is not proof that the website changes caused the increase. The underlying search systems may also have changed, and results may vary by location or session.
Nor does the increase prove that revenue improved.
However, the method provides a more useful starting point than simply asking an AI tool whether it knows the company exists.
The key is consistent measurement combined with commercial context.
The commercial risk is being overlooked before the enquiry
Not every AI search interaction will produce a click.
Not every mention will generate an enquiry.
And not every search query deserves equal attention.
The risk for businesses is therefore not simply that organic traffic may decline.
It is that customers may be forming an understanding of a market, evaluating providers and narrowing their choices through systems that the business is not adequately monitoring.
For specialist firms, professional services and organisations with complex propositions, that deserves attention.
The objective should not be to maximise mentions at any cost.
It should be to ensure that the business can be found, understood and represented accurately when commercially relevant questions arise.