A solo developer sat down this week to write an ordinary comparison post for his niche. Instead of skimming what other blogs had written, he did something unusual: he opened every vendor's actual checkout and pricing page.
One tool had a free tier listed across three different Best of 2026 guides. That tier does not exist anymore. A comparison table claimed a product offered 2,000 API calls without mentioning that API access only starts at a 189 dollar monthly plan. He nearly published a billing mistake of his own before catching it on a second read.
The same week, another founder built a benchmarks page from 54 statistics and forced himself to confirm each number inside the original PDF reports. His dry conclusion: a lot of blog posts cite numbers that are not in the source.
What is a citation loop?
A citation loop is secondhand evidence quoting secondhand evidence until nobody remembers where the number came from. Post A cites post B, post B cites a roundup, the roundup once skimmed a pricing page that has since changed. Every link in the chain was cheap to produce, and that is the point: checking the source is the only expensive step, so it is the step everyone skips.
This is also the research most founders run before building something. Google the market, read three roundups, nod at the same recycled figures, feel informed. The market size, the competitor pricing, the churn benchmark: all of it may be three hops away from a page that no longer exists.
Does AI make the loop better or worse?
Worse, and faster. Answer engines are trained and grounded on exactly these posts, so the dead free tier becomes a confident, well-formatted answer. The same week these two builders published their findings, a venture firm reported that an LLM is now the most common way founders say they found the firm. The loop is not a corner of the internet. It is becoming the front door.
Generating a plausible number now costs nothing. Verifying one costs what it always did. When production gets cheap and verification stays expensive, the share of unverified claims in circulation can only go up.
How do you research a market without joining the loop?
Treat any number you did not trace to its source as a rumor with good formatting. For facts about products, the vendor's live pricing page outranks every roundup. For facts about reports, the PDF outranks the post that quotes it. The two builders above did nothing sophisticated. They just opened the originals, and the originals disagreed with the consensus.
For facts about demand, there is no document to open. The primary source is a person who matches your buyer, telling you what they currently do and what they actually paid for. That evidence is expensive for the same reason it is valuable: nobody can copy and paste it into existence, and it does not expire quietly in someone else's archive.
Key takeaways
- Published comparisons and roundups routinely cite dead pricing pages and numbers missing from their own sources.
- AI answer engines amplify the loop by summarizing the summaries, with full confidence.
- Any number that changes your decision deserves a trip to the primary source: the live page, the original PDF.
- For demand there is no primary document. The source is a real buyer describing real behavior, firsthand.
Why this today: both findings, the ghost free tier and the stats missing from their sources, were published independently this week by builders doing ordinary work, which is exactly what makes them representative.
The internet quotes itself. Buyers do not.