← Strategy & Moats
Foundational

Answer Engines and Your Traffic: AI Overviews, AI Mode, ChatGPT and What a PM Should Do

AI summaries in search and chat assistants answer many queries without a click. Learn what the evidence says about lost clicks, which of your queries are exposed, what Google says about special files, and the product moves that protect a business: earn citations, own what an answer cannot replace, and measure by query type.

CONCEPT / STRATEGY & MOATS

TL;DR: Discovery now runs through AI answers at enormous scale: Google reports AI Mode passed a billion monthly users in May 2026, and OpenAI reports more than 1.2 billion weekly ChatGPT users. When an AI summary appears, people click less: Pew measured traditional-result clicks falling from 15% to 8% of visits. Expect losses on queries a summary can fully answer, protect the ones it cannot, and measure by query type rather than in one organic-traffic line. Google says no special files or markup are needed to appear in its AI features; an llms.txt file is an informal proposal, not a ranking lever.

What the evidence says

DIAGRAM
Google searches with and without an AI summary no AI summary AI summary shown Clicked a traditional result 15% 8% Clicked a link in the summary 1% Ended the browsing session 16% 26% Share of visits. Pew Research Center, browsing data from March 2025, published 22 July 2025. Bars drawn to scale.
Read diagram description

Pew Research Center browsing data from March 2025. On Google searches without an AI summary, users clicked a traditional result on 15% of visits and ended their browsing session on 16%. With an AI summary, they clicked a traditional result on 8% of visits, clicked a link inside the summary on 1%, and ended the session on 26%.

Pew's measurement used real browsing behaviour from March 2025. Two later studies point the same way with different methods. Ahrefs, an SEO vendor, found in February 2026 that an AI Overview correlated with a 58% lower click-through rate for the top-ranking page across 300,000 keywords; that is a vendor correlation, not an experiment. A 2026 working paper by Agarwal and Sen randomly hid AI Overviews for about a thousand US desktop Chrome users, and a summary from the authors' institute reports organic clicks cut by around 40%. The magnitudes differ; the direction does not.

Google's position is that the clicks it does send from AI features are "higher quality (meaning, users are more likely to spend more time on the site)". That is a company claim, and it can be true at the same time as a fall in volume.

The business stakes are real. Chegg told investors in February 2025 that its non-subscriber traffic was down 49% in January and that it would not need to review strategic alternatives "if Google hadn't launched AI Overviews" (company statement in an SEC filing).

Which of your queries are exposed

Sort your organic queries into three groups before deciding anything.

Query typeExampleWhat an AI answer does
Answerable"What is a context window?"Answers it fully; expect the click to go
Deliberative"Best laptop for video editing under ₹1 lakh"Builds a shortlist; you need to be on it and cited
Doing"Convert this PDF", "book a table", "log in"Still needs your product; clicks hold up best

Estimate exposure by weighting each group's traffic by how much a summary can satisfy it. A content business built on answerable queries faces a structural loss; a product whose pages are mostly doing queries faces a smaller one.

What actually helps

Be the cited source. Assistants cite pages that state a clear, dated, specific claim, especially original data. Publish numbers others quote, keep pages current, and make each page answer one question well. Google's guidance for its AI features is the ordinary SEO guidance: be indexed and eligible for a snippet.

Do not buy myths. Google's Search Central page says there are "no additional requirements to appear in AI Overviews or AI Mode" and that you "don't need to create new machine readable files, AI text files, or markup". The llms.txt proposal describes a file that helps agents use a website, and its authors report adoption by thousands of sites and documentation platforms. It may help coding agents read your API docs. It is not a Google ranking signal.

Choose controls with eyes open. nosnippet, data-nosnippet, max-snippet and noindex limit how your content appears in AI features, and they limit ordinary snippets too. Blocking is a trade of visibility for scarcity, worth it mainly when the content is the product people pay for.

Own what a summary cannot replace. Tools, calculators, accounts, communities, proprietary data and workflows survive because the user has to come to you to use them. Shift investment from pages that explain to products that do.

Diversify distribution. Email lists, apps, partnerships and direct brand demand are channels no answer engine intermediates.

How to measure it

Track impressions and clicks separately by query group in your search console, since a stable impression count with falling clicks is the AI-answer signature. Track referral visits from assistants as their own channel. Sample a fixed set of prompts in the major assistants each month and record whether you are cited. Watch branded search volume, which captures demand created upstream of any click. Read a falling organic line through diagnosing metric drops before blaming AI: seasonality, a site change or a core update can look identical.

In the interview

Expect "our organic traffic fell 30% since AI Overviews; what do we do?" or, from the other side, "how should our assistant treat publishers?". A weak answer proposes "GEO" tricks or an llms.txt file. A Hire answer works on content quality and citations. A Strong Hire answer segments queries by exposure, quotes the evidence with its limits, separates measurement of impressions, clicks and citations, invests in what a summary cannot replace, and treats snippet controls as a deliberate trade.

Next: product strategy and distribution and platform risk and lock-in.

RELATED CONCEPTS
PRACTISE THIS CONCEPT