The Post-Text Economy and the Disintegration of Authenticity
Text is losing ground to video, people are not simply handing financial decisions over to an AI, and machines read your site differently to how a customer does. This page explains what that means for the way you organise yourself online. This is an analysis, not research: there is no dataset behind it and we have never carried out our own measurement for it. On 18 August 2026 every figure was removed from this page because we could not show the source file, and on 19 August 2026 the research format was removed as well. What used to be here and why it is gone comes first, below.
What has been removed from this page, and why it is now called an analysis
On 18 August 2026 we checked every figure on this site against its source, including our own figures. For this page that trail ran cold. It carried around thirty-six figures, presented as the outcome of our own research: a sample of 4,302 consumers and 840 B2B decision-makers, an eye-tracking group of 53 students, a Turing test among 77 Dutch respondents, a threshold amount of €60.07 down to the cent, and a series of percentages on platforms, developers and ad tests. We could find no file, no table, no export and no note to support any of it, and no trace externally either.
Figures like that call for a source file you can open
There isn’t one, so the figures are gone. The same goes for the line that we spent €12,000 on ad budget and analysed 1.2 million impressions: that is an accounting claim, and there is no printout behind it.
On 19 August 2026 a second step followed, because removing the figures alone was not enough
This page still carried the shape of research: a number, a registration mark, the word "sector study" and a block describing a method with a survey among consumers and business decision-makers plus eye-tracking among students. That survey and that eye-tracking were never carried out. Describing the method of research that does not exist is, even without figures, a claim we cannot support, so everything that pointed to that apparatus is gone. From now on this piece is an analysis: our own argument about where text, authenticity and machine readability are heading, without a dataset of our own and without a measurement of our own.
What remains is that argument, and you can test it yourself: in your own scroll map, in your own ad account and on your own timeline
What we can show is in the open data of the National AI Index NL: 2,520 logged answers, complete with the questions, the dates and the invalid rows. Our rule since this round is that a figure comes with a source and a link, or it does not appear at all. Since 19 August there is a second rule alongside it: something is only called research if there is a file behind it that you can download.
Four illusions that no longer hold
The outcome is uncomfortable for anyone who still sees their website mainly as a text product. The assumptions about SEO, authenticity and AI autonomy no longer hold.
Three signals stand out
The TTV idea (Time-To-Video) says that tolerance for text without a visual trigger has become short; exactly how short, we no longer attach a figure of seconds to that. The human debugger largely refuses to hand financial control over to AI. And the B2M paradox shows that smart companies build two faces: visual content for people at the front, pure API data for AI agents at the back.
This page is our own argument, not a research report
This is where a methodology note with sample size and date used to sit; it has been withdrawn because there is no file behind it, and on 19 August 2026 the rest of the research format was removed as well. See DOC.00 above.
The four illusions of 2026
The foundations the internet has run on for twenty years are changing fast. Four stubborn assumptions do not hold up.
The illusion of the reader
Human text consumption is declining, and among the youngest visitors text is increasingly skipped rather than read as a source of information. How large that effect is, we will not say, because we have no measurement for it.The illusion of autonomy
Because of hallucination, AI takes over almost no complex process completely. That forces the human into a permanent, supervising role: the human debugger.The illusion of visual truth
AI slop has destroyed the authenticity premium. The consumer can no longer see the difference and sometimes even prefers the perfect AI image over the real person.The illusion of the search engine
Text counts for less and less with the youngest group, and an AI agent does not read your site as a visitor but as a data source. That makes the traditional SME website less relevant over time.The post-text economy and the fall of text SEO
The shift from text to video is often dismissed in the media as an attention problem among young people. This is where our own eye-tracking measurement among 53 students used to sit, with a retention figure attached; it has been withdrawn, see DOC.00.
The observation that still stands, and one you can repeat yourself on your own site: the youngest visitors increasingly do not read text as a source of information but skip it as a barrier in the way of one. Curious whether the same is true for you? Put a scroll map on your most important page and see how far people get.
The TTV metric and the end of the weblog
That comes with a term we use ourselves: TTV, Time-To-Video. This measures how long a visitor stays on a page before looking for a playable video and, if none appears, leaving. A video that only plays in the background without adding value turned out to have the same impact as a full video.
Gen Z, aged 18 to 29
If a page does not offer a dynamic, visual format within a few seconds, a large share of this group drops off. Three percentages used to sit here; they have been withdrawn.The inversion of Generation X, aged 44 to 60
Where in 2023 it was still assumed that older generations held on to text, in 2026 we instead see a reversed, top-down adoption of video. Two percentages on the media use of B2B decision-makers used to sit here; they have been withdrawn.The end of text ads
We are not adding a decline percentage here, because a figure on the CTR of purely text-based search ads would have to come from our own accounts, and that is not a publicly verifiable measurement. What you can check yourself is in your own Google Ads account: line up the CTR of your search ads over the last three years.From B2C text to B2M data
While the young consumer demands almost nothing but video, something different is happening at the back end. AI systems that autonomously scan the web cannot watch video the way people do. They look for structured data.
A company preparing for 2027 to 2030 therefore has to do two things at once: distract the human with video at the front, and serve the machine with raw API endpoints and business-registration checks at the back. We sum it up like this: the traditional text-based about-us page has become a relic. You build your front end for the human eye and your back end for the autonomous AI agent.
The copilot syndrome and the human debugger
The biggest misconception in business in 2026 is that AI replaces jobs one for one. What does happen: the nature of the work changes drastically. We are shifting from a society of makers to a society of debuggers, while in day-to-day life AI still takes over very little on its own.
The Dutch conservative threshold
Where American markets quickly integrated AI into decision-making, the Dutch market shows more reticence. The Dutch demand efficiency but hate losing control. A threshold amount of €60.07 and a figure of 84 per cent used to sit here, both from a survey whose file we do not have; both have been withdrawn. What remains standing is the pattern: working AI integrations follow the iDEAL principle, the AI does the heavy lifting, but the human still has to press the approval button. We call this the debugger moment.
The denial in the workplace
There is a blind spot in the workplace, and it is clearly visible with or without figures. Three percentages on junior vacancies, senior developers and IT managers used to sit here; they have been withdrawn, see DOC.00.
- The inflow of junior developers is under pressure, because a language model writes better basic code than most newcomers.
- Senior developers generally say that AI can never fully take over their job.
- At the same time, managers are openly considering smaller teams made up mainly of reviewers and debuggers who check the mass production of AI code. The fact that AI does not yet fully take over jobs does not mean it will not do so in time.
The disintegration of visual authenticity
Until halfway through 2024, seeing was believing. In the digital ecosystem of 2026, visual proof has largely lost that value. What percentage of an average timeline consists of AI slop, broken down by platform, used to sit here; those four percentages have been withdrawn, see DOC.00. The observation itself still holds and takes thirty seconds to test: scroll through your own timeline and count how many images you can point to with certainty as made by a human. The human reaction to what is left is not rejection but a form of psychological apathy.
Hindsight bias and the generation gap
Our own visual Turing test among 77 Dutch respondents used to sit here, with four percentages on who did and did not spot the fake. That test and those percentages have been withdrawn. What remains without figures is a pattern well described in the literature on hindsight bias: people recognise manipulated images worse than they think they do, and afterwards say more often that they saw through it than was true beforehand. Anyone wanting to build on authenticity therefore cannot lean on the assumption that the viewer sees the difference.
The creator collapse and the automated brand face
The most disruptive shift is happening in the creator economy. Where the product was once the authentic connection, that model has been overtaken by the scalability of generative AI. A single operator can use AI to manage dozens of photorealistic personas that interact continuously. That hits micro-influencers, OF models and small actors hard.
This section used to be named after the 23-29 formula, and that name was removed from the heading on 19 August 2026
What 23 and 29 meant was never stated anywhere on the page; the explanation was tied to the figures withdrawn below, and once those went, a formula name without a formula was left behind. A name with two numbers in it suggests a calculation, and there is none.
We do not back this up with our own advertising figures
A claim about budget spent and impressions analysed is an accounting claim with a test result attached to it, and that only belongs on the page if there is an invoice and an export behind it that we can show. All of it has been withdrawn.
What can be said about an automated brand face without those figures is this, and it is an observation, not a measurement: a persona that is just short of perfect works better than one that is, because a slight asymmetry or an accent sidesteps the uncanny valley. If you want to know what works for your audience, there is only one way: run two variants side by side yourself and read your own result.
The B2M survival guide for 2026 to 2029
To survive the B2M paradox, business-to-machine, we recommend a plan with four steps.
This plan and the years 2026 to 2029 are our own expectation, not a measurement or an announcement by a third party
Written down and last checked on 25 August 2026.
Shift the text SEO budget
Move a substantial share of your budget into short-form video production for TikTok and Reels. A figure of 80 per cent used to sit here; that was a recommendation with a number that suggested a calculation lay behind it, and there is none.Test your brand face instead of choosing it by formula
A recommendation to replace actors with AI-generated female brand ambassadors for maximum conversion, under the name 23-29 formula, used to sit here. That recommendation is gone. It linked a conversion promise to gender and age, and the only evidence behind it was the ad test withdrawn elsewhere on this page. What remains is the advice to run two variants side by side and read your own result.Build in the human-debugger step
Automate processes, but give the client a real check-and-confirm button that genuinely holds the action back until they press it. You often hear the advice that the illusion of control already builds enough trust; a button that does nothing is a lie in the interface. A button that does something builds the same trust and is true as well.Structure for the machine
Make sure your company is machine readable and develop open APIs, so AI purchasing agents can cryptographically verify your offer.The irony of AI is not that machines imitate people, but that people are starting to prefer the machine over the person. Companies that cling to text, organic humanity and traditional search engines become less relevant digitally. A timeframe of 36 months used to sit here; it came from the same withdrawn research and has been removed.
What this means for your business
In practice, not every business runs the same amount of risk. In a second analysis of ours you can read why an AI answer keeps naming the same short list, and what that means for your position.
Want to see where we do have a file behind the numbers? On our research desk, alongside these two analyses, sits the AI Index: an ongoing measurement of what a language model answers per sector, with the raw data available as CSV and JSON so you can count it yourself.
Want to know how your website is placed in this shift from text to video and from human to machine? Book a call and discuss with us what the B2M paradox concretely means for your business.