Intent collapse happens when different people with different goals receive one merged answer.

Intent collapse describes two related failures: a page that tries to satisfy several conflicting goals at once, and a generated answer that flattens distinct user tasks into one summary. In both cases the differences between people disappear. The term comes from SEO practice and has no fixed academic definition.

Search intent is the goal a search engine reads from a query. It is the small, query-based part of the user intent. Intent collapse happens at the level of search intent, where the person has already been reduced to a category.

Intent cannibalization describes several pages of the same website competing for the same user goal. The site divides its own strength. It looks at goals, where keyword cannibalization looks at phrases.

Also discussed at https://searchengineland.com/guide/keyword-cannibalization

Intent compression describes answer engines that collapse a multi-step search journey into one direct answer without any click. Steps that once needed several pages happen inside one summary. The funnel shrinks to a paragraph.

Described in a Google patent: https://patents.google.com/patent/US11769017B1/en

Intent fracture describes an ambiguous query that splits into many possible sub-intents, so that a search engine shows varied results or offers refinements. It is the opposite movement to collapse. Search systems handle it with diversified results and clarifying options.

Described in a Google patent: https://patents.google.com/patent/US11797626B2/en

An intent mismatch occurs when a page does not match what people expected from the result. They leave, reformulate or return to the result list. A mismatch shows in behaviour long before it shows in rankings.

Search intent drift describes how the preferred content for a query changes over time as expectations and result formats change. A page that matched a query two years ago may no longer match it. Intent has a history.

Retrieval collapse describes generative systems converging on the same few sources, reducing the diversity of what people see. The same pages are cited again and again. Everything outside that small set becomes invisible.

AI Overviews place a generated answer at the top of Google's results and reached more than 2 billion monthly users by July 2025. One summary now answers a query that many different people typed. Whoever needed something other than the summary has to look further, and many do not.

Also discussed at https://blog.google/company-news/inside-google/message-ceo/alphabet-earnings-q2-2025/

If being named inside the answer brought the visit back, a cited page would reach the click rate of a result page without an AI Overview. It reaches about half of it.

Generative engine optimization works on being included and cited in generated answers. Against intent collapse, it aims to be the source for a specific need within the merged answer. Precise, citable passages are its means.

Also discussed at https://arxiv.org/abs/2311.09735

Answer engine optimization works on being the source of direct answers in answer boxes, voice assistants and conversational search. It serves the part of an intent that a short answer can satisfy. The rest of the intent is left for pages.

Keyword cannibalization refers to several pages of a site competing for the same exact phrase. It is the older, phrase-level form of the problem. Intent cannibalization is its successor at the level of goals.

Click-through rate is the share of impressions that lead to a click. It falls when summaries answer queries on the result page. A falling click-through rate can mean satisfied searchers or lost ones, and the number alone cannot tell which.

If the answer box only absorbed clicks that searchers would have skipped anyway, both keyword groups would have lost at the same rate. The group with an AI Overview lost 78 % of its clicks, the group without 49 %.

The expectation gap is the distance between what a result promised and what the landing page delivers. It opens wider when a merged answer sets expectations the page never made. People arrive expecting the summary and find a page.

Buying confidence is a person's readiness to act after judging a page. Clear, consistent information raises it, blurred and conflicting information lowers it. Intent collapse costs confidence before it costs clicks.

Measurement and the person. The usual treatment names the collapse and its cousins. It leaves out how to see it in behaviour, through reformulations and task completion, how needs are layered, what confusion and too many choices do to decisions, and when an answer without a click was in fact enough. Several of the measures this would require have not been defined yet.

The query reformulation rate is how often people rewrite a query within a search. When a merged answer or page served the wrong goal, people reformulate. It is one of the few established signals that makes intent collapse visible.

A complex informational query often contains several distinct needs: a definition, a comparison, a decision, a reassurance. Treated as one informational intent, they end up merged on one page. Breaking them down is the first step against collapse.

A searcher hesitation index would measure the small delays and back-and-forth movements people show when a result does not fit what they need. No established measure of this name exists yet. The behaviour itself can be observed in scrolling, pauses and returns to the result list.

Commercial intent entropy would measure how scattered the purchase goals behind a query are, borrowing the idea of entropy from information theory. The term is not yet an established measure. A query whose buyers want very different things would score high and would suffer most from a single merged answer.

A query disambiguation protocol would be a defined procedure for resolving what an ambiguous query means before answering it, for instance by asking back. Search systems already use clarifying questions and diversified results. A named, standard protocol for this does not exist yet.

An information needs hierarchy would describe how needs build on each other, from basic understanding to a confident decision. Information science has related models, such as Robert Taylor's levels of need. Pages that skip the lower levels lose people before the decision.

Zero-click intent satisfaction describes searches that are fully answered on the result page. Information retrieval calls this good abandonment. Not every lost click is a loss for the searcher, and a publisher has to know which intents are satisfied without a visit.

Also discussed at https://research.google/pubs/good-abandonment-in-mobile-and-pc-internet-search/

Decision paralysis is the difficulty of choosing when faced with too many options, studied as choice overload. Merged answers and pages that offer everything to everyone can produce it. People who cannot choose often choose nothing.

Also discussed at https://thedecisionlab.com/biases/choice-overload-bias

The task completion rate measures whether people achieve the goal they came for. It is a stronger measure than time on page or clicks. Content that keeps people busy without solving their problem looks good until it is measured by completion.

Also discussed at https://www.nngroup.com/articles/success-rate-the-simplest-usability-metric/

Semantic satiation is the psychological effect in which a word repeated many times seems to lose its meaning. Repetitive generated text can produce a related fatigue in readers. Pages full of the same phrasing are read with less and less attention.

A cognitive friction index would measure the mental effort a page imposes through confusing structure, beyond loading speed. Cognitive friction is a known concept in interaction design; an index of that name is not an established measure. Usability tests and task completion show the same effort today.

Epistemic curiosity is the desire to know and understand, which drives people to keep exploring after a first answer. A truncated summary can end exploration too early. Search that satisfies curiosity leaves room to go further.

A merged answer serves the most common goal behind a query and leaves the others unserved. Businesses whose customers want something specific disappear inside the average. The customers remain, and their needs go looking elsewhere.

Separate the goals behind a query, give each its own clear answer or page, and ask visitors which goal brought them. Measure reformulations and completed tasks alongside clicks. Asking back is the idea of Supervised Search: the page splits what the answer engine merged.