PSA: I wasted 6 months training my AI on the wrong data set entirely

So I run a local plumbing company in Spokane and started using an AI writing tool to generate our blog posts last January. I was feeding it generic industry articles and expecting it to write about how to fix leaky faucets (you know, the usual stuff). But after 6 months of flat traffic, my buddy who does SEO for a living asked to see my training data. He pointed out I was using content from national plumbing chains, not local ones. That was my moment - I had been training the AI on national keywords like "fix toilet" instead of "Spokane toilet repair" which is what actual customers search for. Has anyone else accidentally trained their AI content tool on the wrong type of source material without realizing it?

2 comments

2 Comments

wade_hall27d ago

Ha, that's like me spending a month teaching my smart speaker to understand my accent but forgetting I have a sinus infection. I feel your pain man. At least you caught it, I had a buddy who trained his AI on Twitter threads about corporate tax law when he was trying to write children's bedtime stories. The stories came out really detailed about depreciation schedules though. Seriously though, that local keyword thing is sneaky because you think "how to fix a toilet" is the same everywhere but then Google's like "nope, show me your zip code first.

beth_baker6926d ago

Ngl that bedtime story thing is hilarious in a sad way. But honestly I feel that sinus infection thing way too hard, I had one last winter and my phone thought I was speaking in tongues or something.