Zero Click Search: What To Do When The Answer Replaces The Link

The measured version of this story is less dramatic than the headlines and more uncomfortable than the reassurances.

Pew Research Center followed the browsing behaviour of a US panel and found that users who landed on a results page containing an AI summary clicked a traditional result 8% of the time, against 15% for pages without one. Users almost never clicked the sources cited inside the summary. That study covered March 2025 and is worth reading directly: do people click on links in Google AI summaries.

Roughly halved, not eliminated. That is the number to plan against.

The Traffic You Lost Was Mostly Traffic That Never Converted

The queries that generative summaries absorb best are the ones with a short factual answer: definitions, unit conversions, opening hours, “what is X”. Those visits were rarely worth much. The queries that still produce clicks are the ones where the user needs to evaluate, compare, download, or trust somebody.

This is not a consolation prize, it is a planning instruction. Sort your pages by which of those two jobs they do, because the strategy diverges completely.

For a definitional page, the realistic goal is now being the source the summary is built from. For a decision page, the goal is unchanged: be the thing worth clicking.

Write The Definition Page To Be Quoted, Not Visited

If a page exists to answer “what is a GGUF file”, accept that most people will get that answer without arriving. Optimise for being quoted accurately rather than for time on page.

Before:

There has been considerable discussion about quantization formats in recent years. Many practitioners have found that the choice of format can have meaningful implications for both performance and quality, particularly on consumer hardware.

After:

GGUF is the single-file model format used by llama.cpp. It packs weights, tokenizer, and metadata together so a model loads without external config files, which is why most desktop local-AI tools ship GGUF rather than the original PyTorch checkpoints.

The first paragraph cannot be extracted because it contains nothing. The second can be lifted whole and remains true, which is exactly the condition for being cited. We go deeper on that mechanic in how to write so AI search quotes you correctly.

Write The Decision Page To Be Unsummarisable

Some things do not compress. A worked example with your actual numbers, a comparison built on criteria you defined, a screenshot sequence, a downloadable template, a position that costs you something to hold. A summary can mention that these exist. It cannot substitute for them.

The practical test: if an assistant produced a three-sentence version of this page, would the reader be done? If yes, that page is a definition page whether you meant it to be or not. If no, it is a decision page, and the click is still available.

Comparison pages are the clearest case, and they have their own failure modes, which we cover in how to write a comparison page.

Stop Counting Sessions As The Only Outcome

The metric that broke is sessions per query. The metrics that still work are downstream: branded search volume, direct traffic, newsletter signups, demo requests, and whether assistants describe your product in your own framing when asked.

That last one is measurable by hand. Ask four assistants the five questions your buyers ask, once a month, and log what comes back. It takes twenty minutes and tells you more than a rank tracker does now. There is a fuller list in the AI search metrics worth tracking.

Do Not Respond By Padding

The reflex when clicks fall is to make pages longer, add more headings, and bulk out the intro so the page “looks comprehensive”. This makes things worse in both directions. Human readers bounce, and retrieval systems chunk your padding and quote it.

The counter-move is the opposite of the reflex: shorter, denser, more specific, more attributable. A 700-word page where every paragraph carries a fact beats a 2,500-word page that says the same thing at a slower pace.

A Context For Compressing A Bloated Page

A Wrivio Context for this could say:

Rewrite this to be denser without losing information. Delete throat-clearing openings, restatements of the heading, and transitional sentences that carry no fact. Keep every name, date, figure, and source exactly as written. Do not add claims, statistics, or examples that are not in the original. Aim for roughly half the length.

Press Ctrl+Shift+Space, paste one section, and read the diff carefully. Watch specifically for a compression that merged two claims into one sentence and quietly dropped a qualifier, which is the failure mode to catch here.

The Honest Position

Nobody knows what the click rate looks like in three years. The measured decline is real, the citation-only future is not yet here, and every confident projection you read was written by someone selling a service for it.

What is safe to do now is unglamorous: fewer pages, more specific claims, sources attached, dates in the text, and a clear split between the pages that exist to be quoted and the pages that exist to be visited.

Common Questions

How much traffic will I actually lose?

Pew measured a fall from 15% to 8% click rate on results pages showing an AI summary in March 2025. Your exposure depends on how many of your queries have short factual answers, so segment before you panic.

Should I block AI crawlers to protect my traffic?

Blocking removes you from citation without restoring the click, in most cases. It is a defensible choice for paywalled or licensable archives and a poor one for marketing content, which we walk through in the crawler post.

Do AI summaries cite the top-ranking pages?

Increasingly not. Overlap between cited sources and the organic top ten has fallen substantially since 2025, so rank tracking alone no longer tells you whether you are visible.

Is long-form content dead?

No, but padding is. Length that carries additional facts, examples, and worked detail still performs; length added to hit a word count now costs you in both surfaces.

Download Wrivio for Windows to compress bloated pages down to the parts that carry a fact, with the diff showing exactly what was cut.