Ahrefs Letaido AI marketing platform beats slop

Every article on the Ahrefs Blog now uses AI somewhere in the workflow, and the internal Ahrefs Letaido AI marketing platform can go from research to nearly publish-ready draft in six to twelve minutes.

That speed is real, but the team deliberately refuses to publish at machine cadence.

Content marketer Si Quan Ong laid out the system in an August 28 post reviewed by director of content marketing Ryan Law, and their definition reframes the usual debate.

They argue slop is not em dashes, the word delve, or any stylistic tell, but content published without enough human understanding, judgment, evidence or original contribution to justify attention.

What counts as AI slop?

Slop transfers effort from creator to reader, forcing the reader to decide what is trustworthy.

Polished prose hides the real decisions about angle, evidence and structure that a one-shot prompt already made.

Alphabet Inc. (NASDAQ:GOOGL) does not automatically punish AI content, Ahrefs notes, yet heavily AI-reliant pages tend to lose search performance over time because average quality drops.

How does Ahrefs do the human work before drafting?

Ong says the team answers five questions before any prose is generated: who the reader is, what the promise is, what point of view is being argued, what evidence is required, and what information gain justifies publishing.

AI can help map SERPs or steelman counterarguments, but it does not make those calls.

The model also needs raw material no web scrape can provide, so Ahrefs feeds it interviews, proprietary data, product details and failed experiments.

Where does human judgment re-enter the pipeline?

Ryan Law built the workflow to mirror an editorial process, not a single magical file, with research, gap analysis, outline and draft saved separately.

Writers must pass gates between stages and decide if the idea deserves to exist, if the outline serves the promise, if claims are supported, and if the draft smuggled in unearned certainty.

Ong dictates with Wispr Flow instead of typing, intentionally keeping caveats and half-formed connections in the input so the system organizes his thinking rather than inventing substance.

What does Ahrefs do with time AI saves?

The obvious play is more articles for less cost, and Ahrefs does use automation to publish and update more, like a Data Refresh Hub that saves at least a day per month updating 12 datasets.

The more interesting return is expanding what fits inside an article, from benchmarks on hundreds of thousands of sites to richer interfaces for AI Mode citations and small free tools like its LLMs.txt generator.

Colleague Mateusz pushes this further with a Source of Truth app in Letaido that stores facts and stats, explanations, product details and how-to guides as a searchable library for agents and writers.

Why does every article still need an owner?

AI may execute, but a specific person must own the result, verify claims and be willing to attach their name.

Law reads every word on the Ahrefs blog and can kill a piece that repeats existing rankings, lacks evidence or fails its promise.

That ownership is the check on sunk cost bias that a polished 2,000 word draft creates every six minutes.

Why this matters beyond Ahrefs

Most SEO and marketing teams now face the same choice between volume and value as AI writing costs collapse.

Competitors like Surfer, Scalenut and Semrush are all selling faster production, while Google's helpful content systems reward demonstrable experience and original data.

StartupHub.ai data shows Ahrefs scoring 36/100 against tracked competitors Semrush at 60/100, Similarweb at 64/100, Peec at 63/100, SpyFu at 21/100 and Whisperingowl at 6/100, which frames the pressure to differentiate on depth rather than output.

Tim Soulo, Ahrefs CMO who often voices the company's contrarian marketing philosophy, put it bluntly that the team built Letaido because the AI marketing platform they needed did not exist.

Ong's prescription is simple: use cheap execution to attempt work you could not afford before, not to publish more average pages.

If no human understands, judges, verifies and stands behind what went live, it is slop.