The internet is filling with content nobody wrote. Not in the spooky sense, in the economic one.
The cost of producing a competent, grammatical, entirely generic article has fallen to roughly zero, and the web is responding the way markets respond when something becomes free: with a flood. Ten thousand near-identical explainers on every topic, each one polished, plausible and interchangeable with the other nine thousand nine hundred and ninety-nine.
Here's the contrarian bit: this flood is the best thing to happen to genuinely original publishers in a decade. When mediocrity is free, mediocrity is worthless, and the engines doing the citing know it better than anyone.
Why engines can't cite the flood
Think about the mechanics of an AI answer. The engine has one response and a handful of citation slots, and its entire value rests on those citations adding something.
An article that restates what the model already knows (which is precisely what generic AI-generated content is, by construction) adds nothing worth pointing to. It's derivative of the model's own training; citing it is citing an echo.
What earns a slot is information the model couldn't have generated itself: a number nobody else has published, an observation from inside a market, an experience that actually happened.
The flood doesn't compete for citations. It's filtered before the contest starts.
The trap: cheap content that costs you
The seductive maths of mass generation (a hundred articles for the price of one writer's afternoon) hides the real ledger. Publishing large volumes of interchangeable content doesn't just fail to earn citations; it dilutes the signal of anything good you've published, spends your crawl attention on pages with no chance of mattering, and teaches both readers and machines that your domain is a place where nothing original lives.
The uncomfortable rule of the new era: content volume is no longer evidence of effort, so it's no longer rewarded as effort.
A hundred generic pages compete with the entire internet's generic pages. One page containing something only you know competes with almost nobody.
What "original" actually means for a small business
Originality sounds like a luxury for research houses, but a small business is sitting on more of it than most publishers. Original data: what you observe in your own operation (enquiry patterns, seasonal demand, what your customers ask, what things actually cost in your market).
A mortgage broker who publishes what genuinely changed in their clients' borrowing capacity this year holds material no model can generate. Real expertise: judgements only practice teaches (which problems look alike but aren't, what the textbook answer misses locally).
First-hand specifics: your prices, your process, your suburbs, cases you've handled (anonymised, honestly told).
None of this requires a research budget. It requires noticing that your daily work produces uncitable-anywhere-else information, and writing it down.
The evidence, honestly framed
Research on answer-engine visibility is young, so hold its numbers loosely, but the direction is consistent with the mechanics above.
Published GEO benchmark research found content restructured around citability (statistics, quotable statements, clear sourcing) gained up to 40% in visibility. And a field study of answer-engine optimisation reported a striking multiplier, with framing worth preserving.
1.82×
more AI citations for content optimised for answer engines in a field study: a result its own authors describe as suggestive rather than conclusive
Suggestive, not conclusive: that's the honest label, and we'd rather keep it than inflate it. But notice what both findings reward: not volume, not polish, but substance made legible.
The moat is having something to say and structuring it so machines can find, verify and quote it, not writing more. That's the craft we unpack in how to write content LLMs cite.
The moat, built deliberately
A practical originality programme for a business of any size looks like this. Publish one genuinely original piece a quarter (your own numbers, your own market observations) rather than four generic posts a month.
Attach a real name, credentials and a track record to it, because engines and readers both weigh who is speaking. Make the original core quotable: state the finding plainly, date it, show where it came from.
And resist the flood's gravitational pull entirely: by all means use AI to draft, structure and polish, but never to substitute for having something to say.
In a web where everyone can generate everything, the only content with durable value is the content only you could have written. For businesses with real expertise that's no hardship: for the first time, the game is rigged in their favour.
No: the tool isn't the issue, the input is. AI drafting original material you supplied (your data, your judgements, your cases) produces citable work faster. AI generating content from nothing produces the flood. The test is whether the page contains anything that didn't exist before you made it.
More than most media companies, on your topic. What jobs actually cost in your area, what fails most often and why, what a fair quote includes, seasonal patterns in your work. That's first-hand information nobody else holds, which is the entire definition of citable.
Publishing less garbage doesn't. Search and AI engines reward substance and demonstrated expertise, not cadence for its own sake. A leaner site where every page earns its place typically outperforms a bloated one, and it's far easier to keep accurate, which engines also notice.
Ask the engines your customers' questions monthly and record which sources they name. Watch analytics for assistant referral traffic and for branded searches after publishing something original. It's imperfect measurement, but it separates content that's working from content that's merely existing.