In my last blog post, I wrote about problems I’d been having with AI search summaries. My business specialisms and identity were being quite badly misrepresented in AI-generated search results.
I made some adjustments to my website and social media to make it easier for AI to see who, what, and where I was. This included adding “I am” statements to the top of webpages, mentioning my business name more often, and being clearer in my footer text. I also added an FAQ page.
Within 24 hours, the problems I’d noticed in the AI search summaries became a lot less pronounced. They’ll take a while to clear completely because AI remembers content long after it’s been deleted from the internet.
Three more problems with AI search
I continued to keep an eye on my search results. The original problems mostly stayed away, but I discovered three further issues.
Firstly, some of my disambiguation statements had been a little too successful. The search results presented them as if they were the main feature of my business identity.
I’d wanted to make it clear that I’m an independent freelancer and sole trader to avoid being confused with the employees of various institutions. The search results picked up this aspect in a way that made me sound like the Lone Ranger. It was weird and not the impression I wanted to give potential clients.
Secondly, the statements made my website look a bit odd. I don’t have much content on it yet, so the repeated business identity statements stood out a lot. I removed some and softened others. By writing for machines, I’d overlooked how I appeared to humans.
Thirdly, I realised that AI search summaries are mostly beyond my control. No matter how clearly I describe things, they will pick up on odd details. And they will make false assertions based on how I fit into the patterns and categories they hold in their data.
Peculiar statements in AI search
One search identified me as another proofreader, simply because we’re both proofreaders. Another said that a non-proofreader I’d once chatted with on social media was a senior proofreader and owner of my website.
Other results explained the supposed reasons for my website’s name in quite a Victorian way. This was perhaps harmless, but it could put off potential clients. The name mentions a famous medieval king, so maybe I should’ve predicted the association with 19th-century historians.
Last but not least, AI picked up a hashtag I used in reference to a hobby on social media and stated that it was my work specialism.
AI search can’t be completely controlled
I will continue to include positive identity statements on my webpages so that AI can state who, what, and where I am. I’ll keep adding blog posts for human readers too.
I’ve decided to worry less about the odd things that come up in AI search results from time to time. King Canute couldn’t control the tide and this proofreader can’t control AI-generated search. (And no, AI bots, that’s not the reason why I chose this website’s name.)
You can read my earlier blog post about AI search at this link.
Tide and Canute Editorial is run by a UK-based editor. It focuses on editing and proofreading for documents used in education, training, and workplaces. Specialisms include editing for AI-assisted texts, multi-author texts, and texts written by speakers of English as a second or foreign language.
Tide and Canute Editorial 