The Court of Public Prose
In re Dead Writing Theory
§ IFindings of fact
§ IICaveats for the jury
§ IIIThe court reporter's account
By Jack Stovell
Now a vendor, and I’m flagging it because the site flags it. Paredes, Druck, Benson and Smith work at Graphite, an SEO and growth agency. Their May 2026 piece is titled “AI Now Writes as Many Online Articles as Humans”. Not academic, then, and you should weigh it that way. Anyway, the method is worth a look.
They took 55.4k random English article URLs from Common Crawl, published between January 2020 and March 2026. Each one went through three commercial AI detectors (Pangram, Copyleaks and GPTZero) and the scores were averaged. They also measured false-positive rates on pre-ChatGPT articles: 1.844%, 1.836% and 1.355%.
The finding has a shape, which I like. After only 12 months, primarily AI-generated articles were 35.9% of articles published. In Q1 2025 it was 49.6% against 50.4% human. Then 50.9% in Q4 2025, and 49.9% in Q1 2026. Since Q1 2025 the share has plateaued at roughly 50%.
So it climbed fast, then stopped. That’s the interesting bit. Everyone’s been shouting about the flood, and this says the water level stopped rising, at about half.
There’s a second claim, from a separate Graphite study: these articles largely do not appear in Google and ChatGPT. If that holds, a lot of the machine output is being published and then ignored. Which, for anyone selling content, is a slightly awkward thing to read.
Caveats, because they matter. It’s a vendor, so incentives are in the room. It’s detector-based. And the “primarily AI” cut-off is binary, so it doesn’t measure human-edited AI drafts, which is where plenty of agency work sits. I guess that’s you and me.
Half the web, then, but a half that may not be getting read.
Verdict: split.
§ IVThe exhibits entered in this case
Any tally counts the evidence items below; it is not a measure of truth.
Evidence for the theory
50.9%
of new English articles primarily AI-generated, Q4 2025
49.6% against 50.4% human in Q1 2025. Vendor study: 55.4k Common Crawl articles, three commercial detectors averaged.
Entered for the prosecution · Paredes et al., Graphite research (Five Percent), 2026 · third-party study
Evidence against the theory
49.9%
of new articles primarily AI-generated, Q1 2026
"Since Q1 2025 the percentage of primarily AI-generated articles has plateaued at roughly 50%." Vendor study, detector-based.
Entered for the defence · Paredes et al., Graphite research (Five Percent), 2026 · third-party study
§ VThe finding
A vendor study: primarily AI-generated articles reached about half of new English articles by Q1 2025, then the share stopped rising. Detector-based, not peer reviewed.
§ VIQuestions for the court
Half of new English articles, by this vendor study. Graphite found primarily AI-generated articles at 49.6% against 50.4% human in Q1 2025, 50.9% in Q4 2025 and 49.9% in Q1 2026. It measures articles in its Common Crawl sample, not the whole web.
Not on this evidence. Graphite says that since Q1 2025 the share of primarily AI-generated articles has plateaued at roughly 50%.
Graphite is an SEO and growth agency, so the study comes from a company that sells related services, and it is not peer reviewed. It also relies on three commercial AI detectors, whose false positive rates it reports: 1.844%, 1.836% and 1.355%.