You Do Not Choose What You Get Cited For

01Seventeen hundred citations, almost none for the product

I opened the citation report on one of my sites expecting to feel good about it. The headline number was excellent, better than anything else I run. Then I read the queries underneath it and realised that the machines had been recommending the site for a quarter of a year for questions I had never once written a page about.

You get cited for the question you answer best, not the one you meant to answer.

What I thought the site was

It has a category, it has a product, and every page on it was written to win the obvious questions in that category. Which tool is best, how do you use it, what does it cost. Those are the queries I researched, those are the pages I built, and those are the ones I checked in the dashboard every week for months. I spent a quarter measuring the questions I had chosen and none of the ones I was winning.

What the machines thought it was

Something much smaller and much more useful. Buried inside the product is a step that solves an unglamorous practical problem, the kind somebody types at eleven at night with a deadline. That step is what got quoted. Not the product, not the category, not a single one of the pages I agonised over. The most cited thing I have ever built is a feature I would have described as a detail.

The mechanism stops being mysterious the moment you look at the pages. That step had one of its own, with a plain title, the answer in the opening line, and no marketing wrapped around it, because I had written it in twenty minutes and never gone back to dress it up. Everything I laboured over was positioning. The one thing I dashed off was an answer, and an answer is the only thing a machine can use.

02The two lists, side by side

Here is the gap, with the shape of each list rather than the exact split, because I have the totals and the themes and I am not going to invent a breakdown I cannot show you.

The number underneath the number

Look at the three figures at the bottom rather than the panels. 1,712 citations against 111 clicks is the whole argument of this blog in one line. The machines used the site sixteen times for every person who arrived on it. I was being recommended far more often than I was being visited, and only one of those two things showed up in a report I read.

03Stop arguing with the result

The instinct on discovering this is to correct it, to push harder on the category pages until the machines start citing the thing you meant. That instinct is expensive and I followed it for about three weeks before giving up on it.

Go and look before you plan anything

Almost nobody reads their citation queries, and the ones who do usually read them for the brand name and stop there. Read the whole list, including the questions that look irrelevant, because the irrelevant ones are where this hides. The monthly audit catches what a model says about you. This is the other half, what a model says you are for.

Then build the thing they already want

Once you know the question you are actually winning, the cheapest growth available is a page that answers it properly, because you are already the answer and nobody has written it down yet. That is the same shape as the small checker that quietly became a third of a product's impressions. The market tells you what you are. Arguing costs a quarter and changes nothing.

None of this means abandon the product. It means the doorway is not where you put it, and a doorway with seventeen hundred citations on it is worth a sign. Two days ago I published a month of my own flat numbers, so I am not claiming this is easy to see from inside. It took me a quarter and an accident.