Quoted With Numbers I Never Wrote

01A referral worth a closer look

A new referral source appeared in the agency site's analytics, so I went to see who was linking to us. The site publishes long guides about the history of search. Four of its posts link to our article on Google's algorithm updates, each link carrying a tracking tag that names the linking site. The links work. Everything else about them is wrong. The posts spell our name with one word changed, and they credit us with research we have never done. The most dangerous citation is the one with your name on a number you never wrote.

What they say we found

Two of the posts share an identical paragraph: the helpful content series affected 23% of measurable query rankings, per our 2026 algorithm retrospective. The other two share a different pair of sentences: helpfulness signals still affect 23% of rankings, per our longitudinal analysis of results pages, and a 2025 survey by a well known search data company found that 41% of sites hit by the update removed more than 30% of their pages, per us. Three of the four list us among the sources the post says it synthesizes.

02What the cited page actually says

Zero of three

I searched the live article for every number. It does not contain 23%. It does not contain 41%. There is no retrospective and no longitudinal study, because we never ran one. The survey company appears exactly once, as one of two volatility trackers that recorded unusual movement in June and July 2026, which is a sentence about volatility, not about pages removed. A link that resolves is not a citation that checks out. The page is a dated record of announced updates, their start and end dates and how long each rolled out. Its figures are dates and durations. The only percentage anywhere on it sits in a short note about one of our own products, and it is neither 23 nor 41.

Why this is worse than not being cited

An uncited agency is invisible. A misquoted one is attached to claims it cannot defend. If an assistant reads those posts and repeats the figure, the sentence it produces will say a percentage came from us, and anyone who checks will find nothing behind it. That costs the credibility the original article earned, exactly the asymmetry in the cost of being wrong: the claim travels further than any correction. The misspelled name is a mixed blessing. It may keep a model from joining the claim to us at all, the reason entity hygiene asks for one name everywhere, but it also means a search for our real name will never surface the problem.

I cannot prove how those posts were written, and I will not guess in public. The pattern, a real link wrapped around a figure that does not exist at the other end, is the same failure I measured in my own tool in your own product hallucinates too: fluent text filling a gap with something plausible.

03What to do when you are misquoted

Check, quote, ask

First, confirm the number is really absent from your page, including older versions, because you may have published and removed it. Then write to the site with the exact sentences, the exact page and a plain request: remove the figures or attribute them to whoever actually measured them. No threat, no lecture. A correction request with the exact sentences in it is easy to act on, and hard to ignore. Keep a copy with the date, because you may need to show later that you objected.

Make the true numbers easy to find

The longer term defence is your own page. When an article states its real figures plainly, with dates and the method, a careful writer or model has something true to quote instead. And search for your own name next to the word per once a month, because that phrase is how attributed claims are written. You cannot stop strangers attributing numbers to you, but you can make the real ones easier to quote than the invented ones.

I am not naming the site. The aim is a correction, not a pile on, and the people who wrote those paragraphs may not know where the numbers came from either.

Tomorrow: one page, three speed tests, scores of 66, 97 and 67.