We're in the prove-it age
Where we came from
In 1900, France had fewer than 3,000 cars on the road. Michelin printed 35,000 copies of a guide telling drivers where to find petrol, mechanics, toilets and a bed, and gave every one of them away free.
Two decades later — in a story Michelin still tells about itself — André Michelin found his guides propping up a workbench in a tyre shop. On the principle that "man only truly respects what he pays for," he pulled the advertising, added restaurants and started charging seven francs.
A century on, a 1904 edition sold at auction for €47,000.
Gloss decays. Usefulness compounds.
The web then broke the assumption that readers read. Jakob Nielsen's 1997 study found 79% of users scanned every new page while only 16% read word by word. A decade of follow-up sharpened it: on an average page, users have time to read at most 28% of the words, and 20% is likelier. The underlying Morkes and Nielsen experiment quantified the fix — concise writing improved measured usability by 58%, scannable by 47%, objective by 27%, and all three combined by 124%. Hooks, summaries and pull-quotes were not stylistic preferences. They were a rational response to measured behaviour.
The lie we built on top of it
Somewhere along the way, "people scan" curdled into "people cannot concentrate." The eight-second attention span — supposedly Microsoft's finding, supposedly shorter than a goldfish's nine — is not real. The BBC traced it: the figure never came from Microsoft's research, and its sourcing chain rests partly on an analytics note about 25 people leaving websites in 2008. Microsoft's actual study mentions neither eight seconds nor goldfish. Dr Gemma Briggs of the Open University is blunt about the premise: "I don't think that's true at all," noting that an average attention span is meaningless because attention is task-dependent. There is no evidence human attention spans are shrinking.
The market agrees. Business podcasts hold listeners for 25 to 40 minutes, flagship shows average well over an hour, and narrative history formats run past four hours across twenty-hour series. On YouTube, 40 to 60 minute episodes now drive close to half of watch time and revenue. If attention is changing, it is becoming more demanding, not shorter.
Where we are
Both things are true at once. People scan, and people go deep — they simply refuse to do the second before the first has earned it. That is what the three levels describe. Provoke earns three seconds. Persuade earns ninety. Position earns as long as the narrative can hold.
Provocation is not a soft art. A 2024 study in the Journal of Business Research analysed 16,215 YouTube video covers and found that strong sentiment, positive or negative, drives more views — while neutral thumbnails underperform. The entry point is a measurable variable, which is why the most disciplined creators over-invest there. MrBeast puts it plainly: "I'm probably the most extreme with thumbnails. We are very obsessed over it." But the same discipline sets the limit — misleading thumbnails win the click and destroy retention and channel trust. Provocative and provable, or not at all.
Where it points
The burden of proof has shifted to the publisher. As synthetic content becomes indistinguishable from the real thing, the onus falls on brands to demonstrate what is genuine. That is now codified. Article 50 of the EU AI Act became enforceable on 2 August 2026, requiring disclosure and machine-readable marking of AI-generated content, with penalties reaching €15 million or 3% of worldwide turnover; systems already on the market have until 2 December 2026 to comply. Google moved the same way in February 2026, releasing its first dedicated Discover core update to reduce sensational content and clickbait while surfacing in-depth, original work from sites with demonstrated expertise. Provenance infrastructure is arriving alongside, through C2PA content credentials that travel with a file and record its origin and edits.
Disclosure alone will not carry you. A 2026 systematic review of consumer responses to AI-generated marketing content found that disclosure activates persuasion knowledge and erodes trust across contexts — meaning transparency creates a deficit that only evidence can offset. Anyone can generate it now. Excellent is what you can prove.
Provoke. Persuade. Position.
P3 is the content discipline. POET (getpoet.ai) is the channel discipline — where distinctive discoverability is earned. The R2 Framework (buildr2.com) is where the result is measured.