Self-Promotional Content material Works—Till It Backfires (AI website positioning Experiment)

When self-promotional content material labored

The clearest end result got here from Ahrefs Evolve, our new website positioning & AEO convention.

As a substitute of trying solely at whole model mentions, I checked out locations the place a particular AI assistant, answering a particular question, had not talked about the model in any respect in the course of the Feb 7–12 baseline.

After the primary pages went reside on Feb 13, Evolve broke into 72 of these beforehand empty slots. Of the 72 new Evolve mentions in prompt-engine slots that had been empty earlier than publishing, 82% appeared in solutions that cited certainly one of our pages.

That’s the cleanest proof that the pages did one thing. The model went from absent to current, and 4 occasions out of 5, our web page was the cited supply.

Selling Ahrefs Model Radar informed the other story.

Ahrefs was already a longtime model, so AI assistants named it throughout many related searches with or with out our assist. When Ahrefs broke right into a beforehand empty slot, solely 6% of these new mentions got here from solutions that cited certainly one of our pages. The opposite 94% got here from third-party content material.

So the self-promotional pages appeared to matter far more for the brand new model than the established one. For Ahrefs Evolve, the content material usually acted just like the bridge between “not talked about” and “talked about.” For Ahrefs Model Radar, AI already had loads of different sources to work from.

Copilot was the AI assistant with essentially the most mentions. Evolve went from basically zero earlier than launch to 39% of Copilot solutions in March, 57% in April, and 65% in Might. Gemini picked it up too, however much less steadily. Perplexity barely moved. ChatGPT by no means named Evolve for this question at all.

And only for the document, listed here are a few of the pages I created to advertise Model Radar and Ahrefs Evolve: