Why guest posts get rejected
If you have ever asked “why was my guest post rejected?” after a silent editor or a one-line “not a fit” reply, the answer is usually not the topic. Editors at sites that accept guest posts process dozens of pitches a week, and they have learned to screen drafts in under a minute. That screen has three parts, and they run in a fixed order.
First comes the AI check. Editors do not run your draft through a detector — they skim for the writing patterns that unedited AI text produces, because those patterns predict the rest of the article. Second comes the link check. A guest post almost always carries a link, and editors know it; what they are judging is whether the article would still make sense if the link were removed. Third comes the substance check: does any paragraph contain something the editor's audience could not get from the first page of Google?
A draft that fails the first screen never gets the second one. That is why a well-researched article with three AI-sounding paragraphs gets rejected while a thinner but cleaner one gets published — the editor never read far enough to notice the research.
The AI tells editors check for
This checker works as an AI content checker for guest posts specifically: it looks for the tells editors actually act on, not statistical perplexity scores. The big four:
- Em-dash rhythm. Language models lean on the em dash as a default pause — several per paragraph, often in pairs. Human editorial writing uses commas, parentheses, and full stops, reserving the dash for genuine emphasis. More than a couple of em dashes per thousand words is the single fastest tell.
- Cliché vocabulary. “Delve”, “in today's fast-paced world”, “unlock your potential”, “seamless”, “game-changer”, “navigate the landscape”. None of these words is wrong alone; clustered, they mark a draft nobody edited. Our engine tracks 165+ of these patterns because we strip them from our own output.
- The moralizing conclusion. AI drafts end the same way: a wrap-up that restates the article and adds an uplifting lesson — “In conclusion, X is a journey, not a destination.” Editors read conclusions first precisely because this tell is so reliable.
- Uniform rhythm and zero evidence. Sentences of near-identical length, paragraphs that open with “Additionally” or “Moreover”, and claims with no numbers, no tools, no examples, and no first-hand experience. Human writing is uneven because thinking is uneven.
How to fix a rejected guest post
Work the three screens in the same order the editor does. Start with the mechanical pass: cut the em dashes down to one or two per article, replace the cliché phrases with plain verbs, and delete the conclusion's last paragraph — almost every AI-flavored draft improves when the moral of the story goes.
Then fix the link. One contextual link to the promoted domain, with anchor text that reads like a reference (“client reporting dashboard”), not a keyword (“best client reporting software 2026”). If the same anchor appears twice, or the paragraph around the link would collapse without it, editors read the whole article as a delivery vehicle.
Finally, add evidence the editor's readers cannot get elsewhere: a number you measured, a failure you hit, a comparison you actually ran. One concrete paragraph buys more acceptance odds than five hundred words of competent generalities. Re-run the checker after each pass — the score moves with the same signals editors weigh, so you can see when the draft crosses from “risky” to “likely accepted” before you pitch it.
Writing guest posts at volume? See how ProofWrite builds them around your client's link so the draft passes these checks the first time.