AI Slop Pays on X — and Still Won’t Rank on Google

If you have scrolled X lately, you have met the genre: a wronged mother, a courtroom, a nine-year-old walking in with evidence. WIRED’s Steven Levy went looking for who actually writes these threads, and the answer was a stranger with a chatbot and a payout schedule. For anyone who publishes content for a living, the stories are not the interesting part — the price is.

“The first post in a story thread he posted earlier this week, presumably narrated by a second wife abused by her in-laws, notched 1.5 million views in two days. X pays him every two weeks as part of its Creative Revenue Sharing program. BIGBEN says the paychecks are usually between $500 and $700. Others report earning thousands of dollars in each cycle.”

— Steven Levy, WIRED

Our take

There are two content economies operating right now and they pay for opposite behavior. The engagement economy — X’s revenue share, Facebook bonus programs, Shorts funds — pays for impressions. The intent economy — search, maps, AI answer engines, your inbox — pays for being the source someone trusts at the moment they are ready to spend money. AI slop is extraordinarily good at the first and structurally bad at the second. Most of the panic about it comes from people who have not noticed those are different markets.

Run the arithmetic inside Levy’s own reporting. A creator clears $500 to $700 every two weeks. X requires a minimum of five million impressions in a rolling three-month window just to qualify for the program. Divide it out and the slop economy is monetizing at roughly a dime per thousand impressions. We bill more than that to answer a phone. No legitimate business can win a game priced that low, and the useful news is that no legitimate business needs to play it.

The second-order effect Levy does not chase is that this is a platform-policy time bomb, not a culture story. X has already intervened three separate times: the May 2025 lawsuit against creators allegedly gaming payouts with bot-driven comments, the April move to cut aggregator payments, and the July 16 removal of millions of posts that recycled other people’s work. Every revenue-share program that floods eventually gets repriced. If your distribution plan depends on a payout formula somebody else controls, you are renting land on a floodplain.

On the search side the evidence is already in, and it does not say what most people assume. Graphite’s analysis of tens of thousands of crawled URLs found that AI-written articles crossed the halfway mark of new articles published online back in late 2024 — and that those same articles largely do not appear in Google or ChatGPT results. Volume went up. Visibility did not. Two different curves.

That tracks with how Google actually wrote its rules. The spam policy does not say “no AI.” It defines scaled content abuse as pages “generated for the primary purpose of manipulating search rankings and not helping users.” The trigger is purpose and value, not tooling. You can draft with a model all day. You cannot ship forty interchangeable posts a month and expect the algorithm to reward the effort.

Here is where we part ways with Levy. He calls these melodramas “fairly harmless” next to rage bait, misinformation and stolen content. From an agency seat they are not harmless, they are inflationary. Every viral fabrication bids up the cost of the same attention our clients are trying to buy, and programmatic display drops real brand logos next to invented courtroom scenes. Harmless to the reader, possibly. Expensive to the advertiser, definitely.

What we actually do across the client sites we run is unglamorous. A model handles research, outlines, first drafts and cleanup. Then a human has to add the one thing the model cannot possess: the job we ran last month, the number we actually quoted, the thing that broke and what it cost to fix. An AI-only post costs a few dollars in tokens and twenty minutes. A post built on a real operator interview costs two to three hours. Only the second one earns a link, a citation, or a phone call, and we have never once seen that ratio invert.

Levy closes on a “hollow feeling.” For a business, hollow has a number attached to it. If your blog reads like the courtroom thread, your quote form reads like a scam too, and the reader who bounces will never tell you why.

What this means for your business

• Audit your last 90 days of published pages against Google’s own language. If you cannot name what a page adds that a competitor’s page does not, rewrite it or unpublish it. Budget about an hour per twenty pages and do the worst category first.
• Retire impressions as a reporting metric. Replace it with cost per qualified conversation. A dime per thousand views is what slop is worth; a booked estimate is worth three figures. Report the second number to your owner and the arguments about content volume end quickly.
• Put one unrepeatable fact in every piece you publish. A price, a date, a permit number, a job you finished on Boulder Highway. A model can imitate your tone. It cannot imitate your invoices.
• Pull your display placement report this month. Exclude the content farms your ads are landing on. This takes twenty minutes and is the cheapest brand-safety work available to a small advertiser.
• Decide your AI disclosure posture before someone decides it for you. Regulators are already moving on this — see our take on the EU’s new AI labeling rules.

The slop is not going away, because for the people making it the unit economics work. Yours are different, and that is the advantage. Read Levy’s original reporting at WIRED.

---