The Messy Middle Didn’t Get Messier. It Got Harder to See.

For a long time, the “messy middle” was a useful idea for SEOs.

It acknowledged something we already knew: people do not move neatly from awareness to consideration and purchase. They explore, compare, hesitate, seek reassurance and return to questions they thought they had already answered.

For many SEO teams, a disproportionate amount of that journey appeared to sit inside Google. Search queries, result pages, reviews and comparison pages gave marketers a partial but commercially useful view of how people evaluated their choices.

The journey was never truly contained. It was simply easier to observe through one dominant interface.

What has changed is not human behaviour. What has changed is where and how that behaviour is shaped.

The old messy middle was easier to observe

It is important not to rewrite history.

Consumers have always explored and evaluated at the same time. They have always used shortcuts, social proof and familiar brands to reduce uncertainty. Google’s original Decoding Decisions research described this movement between exploration and evaluation rather than a simple journey towards purchase.

What made this manageable for SEO was not that Google contained every decision. It was that Google exposed a useful part of the process.

Queries suggested what people wanted to know. Rankings indicated which sources Google considered relevant. Clicks and website behaviour provided at least some evidence of what happened next.

It was incomplete, but comfortably measurable. We do like things we can put in reports.

That comfort is disappearing.

The messy middle now lives across systems

Discovery and evaluation are no longer anchored to one interface.

People encounter ideas in social feeds, videos, newsletters and community discussions. They may refine their understanding by asking ChatGPT, Gemini, Perplexity or another answer system. They may use marketplaces, review platforms or private conversations to compare choices.

Google may then act as a way to verify what they have already heard. In other journeys, it remains the main starting point. Sometimes it barely appears at all.

This is not evidence that Google has suddenly become irrelevant. Nor does it mean that every buyer has abandoned search engines for an AI assistant.

The evidence is considerably less dramatic than some “search everywhere” presentations suggest. As I have discussed elsewhere, AI search has not replaced search engines. Google remains enormously important.

The strategic change is that it can no longer be treated as the only environment in which preferences are formed, checked or reinforced.

The messy middle did not explode. It dispersed.

Google is also changing the journey from within

One weakness in many “search everywhere” arguments is that they treat Google as a static reference point: the old centre from which users are gradually drifting away.

Google itself is changing.

AI Overviews synthesise information directly within the results. Instead of seeing only titles, snippets and links, a user may first encounter an AI-generated interpretation of the subject.

That interpretation can introduce options, summarise differences and frame the terms of the decision before the user visits an external source.

There is already evidence that this affects clicking behaviour. A Pew Research study of US browsing activity found that users clicked a traditional result on 8% of visits when an AI summary appeared, compared with 15% on pages without one. Only 1% clicked a source from within the summary itself.

The study covered US users during March 2025, so it does not establish the same effect for every market, subject or query. It does, however, make the old visibility-to-click assumption rather harder to defend.

Part of the messy middle now happens inside the search results themselves.

From navigation tool to decision layer

AI Mode pushes this further.

Google now allows users to move from an AI Overview into a contextual conversation, asking follow-up questions without beginning another conventional search. Google describes this as a more fluid path between a quick summary and deeper exploration.

Users can refine requirements, compare alternatives and ask increasingly specific questions without encountering a traditional list of results at every step.

The familiar sequence of a query, a set of results, a click and subsequent evaluation therefore becomes less reliable.

This does not mean that nobody clicks. Google still includes and promotes links, and complex or consequential decisions frequently require deeper research.

It means that clicking is no longer required for every stage of exploration and evaluation.

For SEO, that is a structural change rather than another result-page feature.

• Visibility does not necessarily produce traffic.
• A ranking does not tell us how much influence the result had.
• Indexation does not guarantee inclusion or accurate representation in an AI-generated answer.

Content may influence an answer without producing a visit. Conversely, pages that are vague, weakly evidenced or poorly aligned with the question may never become visible in the generated response.

We cannot observe that selection process reliably enough to turn it into another neat ranking checklist. Someone will try, obviously.

The messy middle is fragmenting in two directions

These changes are happening in two directions at once.

Outward, discovery and evaluation are spreading across social platforms, video, communities, marketplaces, AI assistants and private conversations.

Inward, Google is incorporating more of the exploration and evaluation process into its own interface.

The important point is not that one platform has defeated another. It is that the buyer’s process is becoming less visible to any single organisation.

A user may first encounter an idea in a video, discuss it with an AI assistant, look for objections on Reddit, confirm the brand through Google and eventually buy through a marketplace or direct visit.

Analytics may record only the final fragment.

This is why “people will eventually return to Google and click” is no longer a safe planning assumption, even when Google remains involved somewhere in the journey.

Why this challenges traditional SEO models

SEO matured in an environment where optimisation largely meant influencing retrieval.

If a page could be found, it could compete. If it ranked prominently, it had a reasonable opportunity to attract the visit. That visit could then be measured, attributed and placed into a report.

That logic becomes less reliable when:

• discovery happens without a conventional query;
• evaluation begins before a website visit;
• answers are synthesised from multiple sources;
• and commercially important interactions happen somewhere analytics cannot observe.

This does not make SEO irrelevant.

It makes narrow SEO thinking insufficient.

Optimising only for rankings assumes that the buyer’s journey is still organised around lists of links. Increasingly, it is not.

Adapting means optimising for interpretation

Content must now work in more than one context.

It still needs to satisfy the person who visits the page. But its arguments, evidence and descriptions may also be summarised by search features, answer systems, social platforms and other people repeating what they understood.

This does not require a mysterious new writing style. Nor does it justify chopping every page into tiny “AI-friendly” statements.

It requires the less glamorous disciplines SEO should have valued already:

• clear explanations;
• specific and supportable claims;
• consistent descriptions of the business;
• credible evidence;
• and enough context to prevent a useful statement becoming misleading when separated from the rest of the page.

The objective is not to manipulate a language model. It is to make the organisation easier to understand and harder to misrepresent.

What senior teams should examine

For a senior team, this creates four practical questions:

• Where do buyers first encounter the category, problem or brand?
• Which sources are likely to shape the answers they receive?
• Is the business described consistently across its website and credible third-party sources?
• What evidence helps a buyer or an answer system distinguish its claims from everybody else’s?

This is not a demand to be active on every platform.

Channel sprawl is not a strategy. It is usually what happens when nobody is willing to choose.

The job is to identify the decision points that matter and determine whether the business appears there in a credible and coherent form.

For in-house SEOs, that means interpreting behaviour rather than merely optimising a system.

For agencies, it means offering judgement rather than pretending fragmented and opaque systems can be made certain.

For CMOs, the useful question is no longer simply, “How do we rank?”

Where are decisions being shaped, and do we appear there in a way that helps the buyer trust and understand us?

Measurement must become more honest

Much of this journey cannot be observed directly.

Evaluation happens in private conversations, closed communities, algorithmic feeds and AI interfaces that provide little or no useful referral data. Even when a system cites a brand, it may not send a measurable visit.

That does not make measurement pointless. It means we need to stop asking individual metrics to prove more than they can.

Rankings, clicks, branded searches, direct visits, assisted conversions, mentions and customer research can each reveal part of the picture. None provides a complete account of how the decision was made.

The goal is not perfect attribution. It is enough evidence to make better decisions.

The messy middle is still there

The messy middle has not disappeared, and human decision-making has not suddenly been reinvented by generative AI.

People still explore, compare, hesitate and seek reassurance.

What has changed is our ability to observe that behaviour through one dominant platform and one familiar set of metrics.

SEO still matters, but not because every journey ends with a blue link and a measurable click. It matters because search, in its conventional and AI-powered forms, helps shape how organisations, products and choices are understood.

That makes representation as important as visibility.

It also requires less certainty, fewer universal checklists and a clearer understanding of how people actually choose.

---