The market

We have been telling you the market moved. Here is who measured it.

Every other page here is our reading of where things went. This one is not ours. It is what independent researchers found, with their names and their dates on it, including the findings that make our case harder.

Every figure on this page names the organization that measured it, the study it came from, the sample behind it, and the day it was published. Nothing here is ours.

Sources last checked against their originalsThis date is rendered from the register that holds the figures, not typed onto the page. When an entry passes its recheck date it becomes a finding, and the finding returns on every build until somebody looks again.
What was measured

Your page was read. The person it was read to never arrived.


The assumption underneath every content decision used to be that being found leads to being read. That link is what broke, and it broke measurably.

The first two figures below come from logged browsing rather than a survey, which is why they carry more weight than anything else on this page.

8%When an AI summary appeared above the results, people clicked a link on 8 percent of visits. When it did not, they clicked on 15 percent.Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results. Published 2025-07-22. 900 US adults on the Ipsos KnowledgePanel; 68,879 unique Google searches logged in March 2025, 12,593 of which returned an AI summary.
1%Links inside the AI summary itself were clicked on 1 percent of visits.Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results. Published 2025-07-22. 900 US adults on the Ipsos KnowledgePanel; 68,879 unique Google searches logged in March 2025.
68%In the first four months of 2026, 68 percent of Google searches ended without a click to any website, up from 60 percent in 2024.SparkToro, In 2026, Less than One Third of Google Searches Still Send a Click. Published 2026-06. Similarweb US clickstream panel, January to April 2026, desktop and mobile.
The finding that surprised us

The machines are not rewarding new. They are rewarding kept.


Sample what the engines actually cite, and most of it was updated recently while much less of it was published recently. They are not preferring new pages. They are preferring maintained ones.

A page from four years ago that somebody keeps true outranks a page from last year that nobody has touched. What is being rewarded is not publishing, and it is not volume. It is upkeep, which is the one thing almost nobody can do at scale.

3 in 4Three quarters of the pages AI engines cited had been updated within the last year, while fewer than half had been published within the last year.Seer Interactive, Study: Content Recency's Impact on AI Visibility in 2026. Published 2026-07-24. 7,683 dated pages carrying 47,097 citations across ChatGPT, Gemini and Perplexity, March to June 2026.

See how one change reaches every surface at once

Out of your sight

There is no log of it, because there was no visit to log.


The question is asked where you cannot see it, and the answer is given where you cannot see it. Much of the time nothing on your side is touched to produce it, because the model is answering from what it already holds.

34%Web search was switched on for about a third of ChatGPT queries, so most answers are produced without the model retrieving a live page at all.Semrush, ChatGPT traffic analysis: Insights from 17 months of data. Published 2026-04-07. Over 1 billion lines of US clickstream from a panel of roughly 200 million users, October 2024 to February 2026.
45%Forty-five percent of B2B buyers used generative AI in a recent purchase, mainly to gather information about vendors and products.Gartner, Gartner Sales Survey, Enabling Sellers for AI-Driven Buyer Journeys. Published 2026-03-09. 646 B2B buyers, fielded August to September 2025.
Already happening

Buyers already meet your surfaces disagreeing with each other.


This one is not about machines at all, and it was measured by an analyst firm with nothing to sell in this category. The disagreement between your surfaces is not a tidiness problem you will get to. It is the experience most of your buyers report having.

69%Sixty-nine percent of B2B buyers report finding inconsistencies between what a supplier's website says and what its sellers tell them.Gartner, Gartner Sales Survey, reported as 61% of B2B Buyers Prefer a Rep-Free Buying Experience. Published 2025-06-25. 632 B2B buyers, fielded August to September 2024.
60%+Across eight AI search tools, more than six in ten answers cited their sources incorrectly.Tow Center for Digital Journalism, Columbia University, AI Search Has a Citation Problem. Published 2025-03-06. 1,600 queries across eight generative search tools, built from excerpts of ten articles each from twenty publishers.
Why the hardest surfaces stopped moving

Deferring the update was arithmetic, not neglect.


The trade-off we describe is not a theory about discipline. It is arithmetic somebody has measured, and in regulated markets it is severe enough to explain the whole pattern on its own.

When that is the price of one update, letting the German page and the PDF wait is the correct call, taken by a competent person, at the old prices.

4+ weeksAmong medtech companies managing commercial content by hand, six in ten wait more than four weeks for approval, and an asset goes through three to five rounds of review.Veeva Systems, 2024 Veeva MedTech Commercial Benchmark. Published 2024-11-21. More than 130 commercial medtech leaders.
HalfIn more than half of organizations surveyed, sales teams use fewer than half of the enablement materials marketing produces for them.Forrester, Portfolio Marketing Survey, reported in A Call For Zero-Waste B2B Marketing. Published 2023. Sample size not disclosed in the free post; the survey is named.
The figure that argues against us

Assistants send almost no traffic. That is the point, not the objection.


If the case were that a new channel is arriving and you should get in front of it, the figure below would end the case, and anyone who follows this market already knows it.

That is not the case. The clicks are not moving somewhere else. They are not happening. What does arrive converts far better than search, which is what a shrinking, higher-intent funnel looks like on the way down.

0.5%AI assistants send about half a percent of the traffic arriving at B2B sites, and the visits that do arrive convert at roughly three times the rate of search.Orbit Media Studios, AI Traffic Conversion Rates: New Research from 97 B2B Websites and 29M Visits. Published 2026-07-22. 97 B2B websites, 29 million visits, analytics data, July 2025 to June 2026.
Being cited is not being visited.
Where we stand

What we can prove, and what we only reason.


Two of the things we say are not measured by anybody, and we would rather tell you than let a citation nearby imply otherwise. Here is every position we hold about the market, and how we hold it.

What we say about the marketHow we hold itWho measured it
A machine reads your pages and answers buyers who never open them.MeasuredPew, SparkToro, Seer Interactive, and 3 more
Whoever's content shifts with the market first wins, and AI is what lets a competitor move that fast.ReasonedNobody has. We hold it on the mechanism.
In regulated content, a claim that has outlived its approval is exposure, not a wording problem.MeasuredVeeva
When your surfaces disagree with each other, the disagreement costs you authority.Part measuredGartner, Columbia Tow Center
What an update costs decides whether it happens, so the hardest surfaces are the first to stop moving.Part measuredVeeva, Forrester
The question and the answer both happen out of your sight, so the damage arrives as lost outcomes and never as feedback.MeasuredPew, Gartner, Semrush, and 1 more
The gaps, named

Nobody has published a measured figure for how far behind a localized site runs. Not one research firm, not one translation vendor. The problem is discussed constantly and quantified never, so we say it as a mechanism and never as a number.

Nobody has measured how fast a company can get a changed message onto its own surfaces, or what that speed is worth. The rigorous work on speed is about building products, and the famous figure about a late launch comes from an illustrative model in a book from 1991. We hold this one on the mechanism, and we will keep saying so.

That buyers meet your surfaces disagreeing is measured. That a machine weighing those same surfaces trusts all of them less is not, by anyone. It follows from how these systems work, and it is reasoning, so we mark it as reasoning.

What we left out

An unverifiable number is an invented number somebody else invented first.


These figures would have helped us. Every one of them is quoted widely in our category, and not one would survive a reader who checked. They are recorded here so they cannot quietly come back.

The figureAttributed toWhy we do not use it
65 percent of B2B content goes unusedSiriusDecisions, since absorbed into ForresterNo traceable primary. The earliest retrievable citation is a 2016 blog attributing it to a research firm that no longer exists, and the underlying report cannot be retrieved. No published method, no sample, no fielding date. It is also a decade old.
Consistent branding increases revenue by 23 to 33 percentLucidpress, now MarqRoughly 200 self-reporting organizations, surveyed by a vendor selling brand-templating software. Using it would cost more credibility than the number could buy, on a page arguing for rigour.
A product six months late to market loses 33 percent of its lifetime profitMcKinseyThere is no study. The figure is the output of an illustrative model in a 1991 book, and one of its authors has publicly disowned its use as an empirical finding.
Zero-click news searches rose from 56 percent to 69 percentSimilarwebWidely repeated, but every citation routes through secondary coverage because the primary report is gated. It is also about news rather than business buying. The zero-click figure we do publish comes from an ungated primary.
89 percent of B2B buyers have adopted generative AIForrester, by third partiesThird-party pages attribute this to Forrester and none of them cites a report, so there is no primary, no sample, and no statement of what was measured. The adoption figure we publish instead is the conservative one, named and sampled.
94 percent of B2B buyers use large language models in the buying journey6senseA real survey with a large disclosed sample, and we still left it out. It measures touching an AI tool anywhere in the journey, which is a weaker statement than it sounds, and the vendor sells into exactly this belief. We use the low, independent number.
The one thing this does not tell you

What the market does is measured. What it says about you is not.

We are asking you to hold your message to a standard, and this page is us holding ours: every figure sourced, every date shown, every gap admitted. The open question is what a machine currently says when somebody asks about you, and that is the one thing we show for free.

Sources last checked