A year ago, when ChatGPT was two weeks old, I wrote some predictions for 2023 and promised to look back at them. I’d rather do that honestly than write a fresh set and quietly forget the old ones, so here goes.
Prediction 1: a flood of content, fast
Half right. The flood happened. Anyone who reads search results for a living has seen the pages: fluent, generic, and identical in structure to the ten pages around them. Some publishers were caught putting AI-written articles out under bylines without much checking, and got embarrassed for it.
The cat and mouse I expected was less clear-cut. Google’s February guidance said it would judge content on quality regardless of how it was made, and the September helpful content update took “written by people” out of its own description. There were plenty of updates this year, and plenty of sites lost traffic, but I haven’t seen good evidence that Google went after AI content as such, as opposed to unhelpful content generally.
Prediction 2: people start asking it instead of searching
Right about the direction, too early to call the size. Microsoft put chat into Bing in February. Google announced the Search Generative Experience at I/O in May, with AI answers above the results. It’s still an experiment and still not available here in the UK as far as I know.
What I can’t tell you is how much search behaviour has actually shifted. Plenty of people use ChatGPT daily, and plenty more tried it once and went back to Google. I suspect 2024 is when we find out whether answers at the top of search actually cut clicks, assuming SGE comes out of Labs.
Prediction 3: useful in the boring places first
Right. The uses that stuck for me are the dull ones: drafts, summaries, rewording, spreadsheet formulas and bits of code. I still wouldn’t swap a good writer for it.
Prediction 4: a lot of policy written in a hurry
Right, and faster than I thought. Within months there were stories of staff at large companies pasting confidential material into ChatGPT, and some firms restricting or banning it. Italy’s data protection authority blocked it for a few weeks in the spring. I’d still bet that most small businesses have no written policy at all.
The usual ones
GA4. Universal Analytics stopped on 1 July, as promised. The switch went about as well as I expected, which is to say plenty of businesses are now looking at GA4 numbers they don’t quite trust. Data in old UA properties is due to disappear next July, so if you haven’t exported it, there’s still time.
Cookies. Chrome is now due to start restricting third-party cookies for 1% of users in early January, with the rest planned for later in 2024. After several delays I said not to relax. I’d say the same again: this one does seem to be happening.
Costs. I expected budgets to be cut or frozen in the first quarter, and closer scrutiny of paid media. That seems to have been true for much of the year, with plenty of businesses preferring short, defined projects to open-ended retainers.
Peak season
I haven’t got data worth quoting, and I’d rather not repeat headline figures I can’t check. My impression from Black Friday is that discounting started earlier and ran longer than ever, with “Black Friday” offers in some inboxes by mid-November. That makes the day itself less meaningful and November as a whole harder to read against previous years, particularly with GA4 counting things differently from UA.
If you’re doing your own review in January, I’d compare November and December as a single block against the same block last year, using your ecommerce platform rather than GA4 for revenue, and look at margin as well as sales.
Next
I’ll write a fresh set of predictions in the new year. One thing already near the top of the list is what AI in search does to organic traffic, once we can see it outside the US.