The Answer Engine: How the Click Economy Ends and What Agencies Publish Instead

For three decades, search worked as a simple transaction: a user typed a query, received a ranked list, and clicked through to evaluate sources. Every business competed for position on that list. The list produced clicks. Clicks produced revenue, and - not incidentally - produced the numbers agencies put on monthly reports.

That transaction is breaking down. Within a year of Google launching AI Overviews, Similarweb found the share of news-related searches ending without a click had risen from 56% to 69%.¹ Pew Research found that when an AI summary appears, users click a result on 8% of visits, against 15% without one.² Meanwhile AI-sourced referral traffic across 19 analysed sites grew 527% between January and May 2025.³ Volume in the new channel is rising sharply; the share of search that produces a click to anybody’s website is falling just as sharply.

For an agency this arrives first as a reporting problem. The sessions line on the client deck goes down, the work has not got worse, and the explanation you have available is not yet a strategy. This page is about the strategy. Every figure on it carries a footnote to its source; anything we could not source has been left out.

Metric	Data point	Source
News-related Google searches ending without a click, May 2025	69%, from 56% in May 2024	Similarweb1
Clicks on a result when an AI summary is shown	8% of visits, against 15% without	Pew Research Center2
AI-sourced traffic growth, January to May 2025	+527%	Previsible3
Organic CTR where an AI Overview appears	1.76% to 0.61%	Seer Interactive4
AI search conversion against Google organic	14.2% against 2.8%	Superprompt5
ChatGPT weekly users, February 2026	900 million	OpenAI6

What the recommendation economy actually is

The recommendation economy is the environment where AI systems, rather than ranked lists, determine which vendors a buyer discovers - delivering one answer with implicit endorsement and no visible alternatives.

A traditional results page shows ten organic results, paid placements, and features competing for attention. A client with budget can buy their way onto the same visible field as a better-optimised rival. A recommendation is singular. When a buyer asks an assistant “which vendors handle contract lifecycle management for a mid-market legal team”, the assistant selects one answer, or at most a short list. No ads. No competing listings alongside. The filtering, evaluation and choosing already happened.

This changes what a client is buying from you. In the click economy they bought exposure and you reported on it. In the recommendation economy they are buying eligibility to be named, which is a harder thing to sell and a more durable thing to own.

The zero-click timeline, briefly

Zero-click search grew from a fringe concern to the dominant outcome through deliberate Google product decisions, then accelerated with AI Overviews.

Featured snippets, knowledge panels and “People Also Ask” had been chipping away at clicks for years before AI Overviews launched in May 2024. In the US, 60% of Google searches ended without a click to the open web in 2024; by the first four months of 2026 it was 68%.⁷

The downstream effect on publishers is the clearest available evidence of scale. Organic click-through on queries with an AI Overview dropped 61%, from 1.76% to 0.61%, between June 2024 and September 2025.⁴ Similarweb tracked organic traffic to news sites falling from a peak above 2.3 billion monthly visits in mid-2024 to under 1.7 billion by mid-2025.¹

If your clients publish in order to earn sessions, that is the curve they are on, and no amount of craft reverses it within the old model.

Why the traffic AI does send behaves differently

AI-referred visitors convert at markedly higher rates than organic search visitors, because they arrive having completed most of their research inside the assistant rather than before starting it.

A Google user types a query, scans, clicks, reads, returns, clicks again, and eventually forms a view. They land on a site near the start of a decision. An AI search user asks a complete question, receives a synthesised recommendation, and - if they click at all - arrives specifically because the assistant cited that source as credible. The visit is a confirmation step, not a research step.

An analysis of 12.3 million visits across 347 businesses between January and September 2025 found AI search traffic converting at 14.2% against Google organic’s 2.8%.⁵ Semrush’s June 2025 research found AI-referred visitors worth 4.4x organic.⁸

Label these as market benchmarks when you put them in a deck. They describe a category, not a promise, and client-specific conversion belongs in the client’s own analytics rather than in a vendor’s claim.

The strategic reading is what matters: volume stops being the primary variable. Position in the answer takes its place.

What AI assistants do with client content

AI systems crawl aggressively to train models and build knowledge bases, and return a small fraction of that value as referral traffic. Cloudflare documents this with the crawl-to-refer ratio: pages crawled per visitor sent back.

In the first week of August 2025, Cloudflare measured Anthropic’s crawlers at nearly 50,000 pages crawled per referred visitor, OpenAI’s at 887:1 and Perplexity’s at 118:1.⁹ Training accounts for nearly 80% of AI bot crawling; crawling triggered by a user’s own action, the kind that can send a visitor back, is under 5%.⁹

The implication for an agency is not to block anything. It is that a client’s content is already being read, processed and synthesised whether or not anyone is managing it. The only open question is whether the synthesis includes their name.

The KPIs that replace the old ones

Rankings, impressions, clicks and CTR measure exposure inside a shrinking click economy. Reporting in this channel needs different instruments, and putting them on the deck early is how an agency gets ahead of the awkward conversation rather than behind it.

Presence in answers. The primary competitive metric: across a defined set of buying questions in the client’s category, how often is the client named, and who is named instead. This is the number that replaces rankings, and it is the one clients understand immediately because it is phrased as a question a customer would ask.

Citation rate. Which specific pages assistants draw from, and how often. A page earning no citations is not being read, regardless of its traffic.

Recommendation frequency. Distinct from mere mention. Being cited as a source is not the same as being explicitly endorsed. This tracks the decision-stage prompts - the ones closest to a purchase.

Framing. How assistants describe the client when they do mention them. Users tend to accept the framing they are given, so the adjectives matter commercially.

What the published work earns. Readership on the pieces you publish, click-throughs from them to the client’s own site, and when the assistants last came back and fetched them. This is the most concrete of the new metrics because it is measured on surfaces you control rather than inferred from someone else’s platform.

KPI	What it measures
Presence in answers	Share of target buying questions where the client is named
Citation rate	Which pages assistants actually draw from
Recommendation frequency	Explicit endorsement, not just mention
Framing	How the client is characterised when named
Published-work performance	Readers, click-throughs to the client site, last assistant fetch

Why a recommendation is worth more than a click

A single recommendation to a high-intent buyer, in an environment with no ads and no visible alternatives, carries higher per-impression value than a search result - even at equal volume.

Consider what surrounds a search result: paid placements above, competitors adjacent, features answering the question without a click, and a back button one tap away. The environment is adversarial and attention is fragmented.

An AI recommendation works differently. The buyer asks in natural language, the assistant synthesises and presents an answer, and the user experiences it as disinterested curation. That perception is not strictly accurate - assistants are shaped by training data, content accessibility and entity prominence - but it is the user’s experience, and it drives behaviour.

What this does to the role of a client’s website

In the click economy, a website was the destination. In the recommendation economy it becomes the evidence an assistant reads before deciding whether to name the client. The visit may never happen.

This is a genuine repositioning, and it is worth walking a client through slowly. A business with excellent products, strong references and a beautifully designed site that blocks AI crawlers, hides content behind JavaScript, or publishes without extractable claims will not appear in answers regardless of how strong it is commercially. Assistants cannot recommend what they cannot read.

Three structural barriers come up repeatedly in client estates:

1. Illegibility - key information embedded in images, PDFs, gated downloads or client-side-rendered components that crawlers cannot reliably reach.
2. Entity ambiguity - inconsistent names, descriptions and facts across the web, making it hard for assistants to assemble a coherent picture of who the business is.
3. No direct-answer content - pages that establish expertise at length but never state a plain, extractable claim an assistant can lift into a response.

The third is the most common and the least technical. It is also the one that requires actual writing rather than configuration, which is why it tends to be the last one addressed.

Building the reporting an agency can defend

Establish a baseline first. Before commissioning anything, measure how a defined set of buying questions is currently answered. Document which produce the client’s name, which produce only competitors, and which produce nothing recognisable in the category. That baseline is what every later report is measured against, and gathering it takes a fraction of the time clients expect.

Model the value per session with the client’s own numbers. Take their conversion rate and average deal value rather than a benchmark. A planning model built from their data survives scrutiny; one built from a vendor’s case study does not.

Track the attribution gap explicitly. Most analytics platforms currently misattribute a large share of AI-influenced traffic as direct. A buyer who receives a recommendation, closes the session, and navigates to the site later is recorded as a direct visitor. AI’s commercial influence is therefore systematically under-reported in standard dashboards. Say so in the report rather than letting the client discover it.

Report on what the published work did. Readership, click-throughs to the client’s site, and assistant fetch activity are directly observable on the pages you publish. They are narrower than a full attribution story and considerably more honest, and they are the material that should shape the next month’s brief.

What actually gets commissioned

The whole argument reduces to one operational question: what do you write next, for which client.

Answering it by intuition is expensive and hard to defend on a renewal call. Answering it by evidence means running a fixed set of buying questions across the assistants, seeing which ones return somebody else’s name, and treating that ranked list as the month’s brief. Then publishing against it, re-asking, and looking at what the published pieces earned in readers and click-throughs - and letting both of those signals steer the following month.

Two inputs, one decision. It is a working process with a named editor’s sign-off in it - yours or your client’s - not an engine that reassigns itself. But it turns a content plan from a proposal into something with evidence behind it, which is the part clients renew.

Frequently asked questions

What is an answer engine? A system that synthesises information and delivers one direct response instead of a list of links. Answer engines make the selection decision on the user’s behalf; search engines delegate it back to the user.

What share of Google searches end without a click? In the US, 68% in the first four months of 2026, up from 60% in 2024.⁷ On news-related searches, Similarweb measured 69% by May 2025, up from 56% a year earlier.¹

How fast is AI search traffic growing? Across 19 analysed sites, AI-sourced sessions grew 527% between January and May 2025.³ Across the world’s top 1,000 websites, AI assistants sent 1.13 billion referral visits in June 2025, up 357% on the year before.¹⁰

Why do AI-referred visitors convert better? They arrive after completing their research inside the assistant, already holding a recommendation. An analysis of 12.3 million visits found 14.2% against Google organic’s 2.8%⁵ - a market benchmark rather than a guaranteed outcome.

What is the crawl-to-referral imbalance? The gap between how aggressively AI bots harvest content and how little traffic they return. In August 2025 Cloudflare measured Anthropic’s crawlers at nearly 50,000 pages per referred visitor, OpenAI’s at 887 and Perplexity’s at 118.⁹

Does being cited in an AI Overview help or hurt clicks? Both, depending on position. Seer Interactive found brands cited inside an AI Overview earn about 35% more organic clicks than uncited brands on the same query, while queries with an AI Overview show 61% lower organic click-through overall.⁴ Being cited is a partial hedge against a decline you cannot otherwise avoid.

---

Published February 2026, figures re-checked September 2026. Found by AI monitors how leading AI assistants answer buying questions, produces the content that addresses the gaps, and reports on what the published work earns - for agency partners, under their brand. See how the partner programme works.

Sources

Every figure above links to the page it was taken from. Checked on 10 September 2026.

Footnotes

1. Similarweb, “Zero-Click Searches”, news-related Google searches, May 2024 to May 2025. ↩ ↩² ↩³ ↩⁴

2. Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results”, browsing data of 900 US adults, March 2025, published 22 July 2025. ↩ ↩²

3. Previsible, “AI traffic is up 527%. SEO is being rewritten.”, 19 GA4 properties, sessions from 17,076 to 107,100 between January and May 2025, published in Search Engine Land, 2025. ↩ ↩² ↩³

4. Seer Interactive, “AIO Impact on Google CTR: September 2025 Update”, 3,119 queries across 42 organisations, June 2024 to September 2025. ↩ ↩² ↩³

5. Superprompt, “AI Search Traffic Converts 5x Better Than Google: 2025 Conversion Data from 12M Visits”, 12.3 million visits across 347 businesses, January to September 2025. A vendor’s own data. ↩ ↩² ↩³

6. OpenAI’s announcement, reported by TechCrunch, “ChatGPT reaches 900M weekly active users”, 27 February 2026. ↩

7. SparkToro, “In 2026, Less than One Third of Google Searches Still Send a Click”, US clickstream data, January to April 2026. ↩ ↩²

8. Semrush, “AI Search Traffic Study”, 9 June 2025. ↩

9. Cloudflare, “A deeper look at AI crawlers: breaking down traffic by purpose and industry”, 28 August 2025. ↩ ↩² ↩³

10. Similarweb, “AI Referral Traffic Winners By Industry”, June 2025, reported by TechCrunch, “AI referrals to top websites were up 357% year-over-year in June, reaching 1.13B”, 25 July 2025. ↩