In May I posted predictions on LinkedIn about how the 2026 World Cup would play out in search. I ended the post with “let’s see how wrong I am in August”. August came and went, and I didn’t come back to it.
I’ve gone back now to mark them. None of them can be marked, right or wrong. That’s a worse result than being wrong, so it gets a post.
What I said
The post split the World Cup SERP into three camps and made a call on each.
- Publishers would “dominate breaking news and editorial” and “lose the quick-answer queries entirely”.
- Brands: “90% of brand World Cup content gets buried.”
- AI answers would win the lookup, while people still needed real sources for anything live.
It also said AI Overviews “already eat 30%+ of clicks on informational queries”. I didn’t give a source for that figure.
Why none of them can be marked
In how to write a claim you can mark wrong I set out four things a checkable claim needs: the date it was made, a fixed sample, a check date, and a result that would prove it false. The May post has the first. It’s missing the other three.
“Lose the quick-answer queries entirely.” I never listed the queries. Without a list written down in May, any query that still sent a publisher a click could be waved away as not a quick-answer query. “Entirely” sounds testable. It isn’t, if you can choose the queries afterwards.
“90% of brand World Cup content gets buried.” Buried means what: off page one, out of the AI Overview, no clicks? 90% of which pages, counted how? Nobody could count it, including me.
“AI wins the lookup.” There’s no definition of winning, so there’s no result that would count as losing.
“August.” A month isn’t a check date. It doesn’t say what gets rerun, or on which day.
What a checkable version would have needed
This isn’t a new prediction. The tournament is over, and a call written after the result can’t be scored. It’s what the May post should have contained.
- A named list of queries, published with the post. For example, a fixed set of “who won” and “what time is kick-off” searches.
- A definition. “Buried” could mean not in the top ten organic results for that list, on the check date.
- A check date: one day, not a month.
- A result that makes it wrong, written down before the result came in.
With those four things, I could have rerun the list on the day and marked each call. Without them, the post was a set of opinions with a deadline attached.
What happens to the post
It stays up on LinkedIn, unedited. It isn’t going on the predictions page, because there’s nothing to score. A row that can’t be marked would only make the board look busier.
Calls I’ve made since are on that page with a check date and a pass/fail rule written before the result. The first two, P1 and P2, are due on 2 January 2027. Wrong ones stay up, in red.
There are no new predictions in this post.