Aensight Journal / Technical SEO
Does Fixing Technical SEO Actually Change What AI Says About You? We Tested It
On day one, Google's AI Overview did not know we existed. Within a week of applying eleven specific fixes, every major AI tool was describing us correctly. Here's the real, week-by-week record.
By Ayushi Agrawal, Founder, Aensight. Published .
On day one, Google's AI Overview did not know we existed. Within a week of applying eleven specific fixes to our own site, every major AI tool was describing us correctly. Here's the real record in between.
These are our own observations from testing our own site. They are self-reported and not independently audited.
This mattered to us for a specific reason: more people are asking AI tools these questions before they ever open a search engine. Why is AI visibility important
Day one: what four AI tools said about us
The day our site went live, we asked four different AI tools the same simple question: what is this company, and what does it do?
ChatGPT didn't know. It asked us which company we meant, since a couple of others share a similar name.
Perplexity did answer, but confidently told us about a completely different company, an unrelated engineering software tool with a similar name. No hesitation, no "I'm not sure." Just a wrong answer, said with full confidence.
Claude searched, came up empty, and said so directly.
Google's AI Overview returned nothing about us at all. Not a mention, not a mix-up, no result.
Day three: Google finds us, and gets it wrong
Two days later, Google's AI Overview returned something for the first time, but mixed us up with a similarly named company, the same kind of mistake Perplexity had already made. Google went from nothing, to something, to wrong, before it ever went from wrong to right.
One week: eleven fixes, applied
About a week after launch, we used our own product on our own site. We fixed the meta description, the H1 heading, the title tag length, and the canonical URL. We added structured data, Open Graph tags, and breadcrumb schema. We corrected the title and H1 to actually include the right keywords, fixed a duplicate meta description, and rewrote key content into clearer question-and-answer sections, the format AI tools pick up and quote most easily.
That's eleven of our own twelve fixes. We skipped one on purpose, LocalBusiness schema, because it's built for businesses with a physical location people walk into. We're not that kind of business, so adding it would have been inaccurate, not thorough.
The Q&A restructuring specifically targets how answer engines read a page, not just whether they can crawl it. Making your brand legible to answer engines
A study out of Princeton, published in 2024, found that structural changes to a page, clearer facts, better formatting, could shift how often AI tools cited it by as much as 40 percent. That's the category of change we made: structure and stated fact, not persuasive copy. Google's AI Overview started describing us correctly that same week, and has continued to since. (Aggarwal et al., KDD 2024)
Where it stands today
Running the same test now, Google, ChatGPT, Perplexity, and Claude all correctly describe who we are and what we do.
It's not perfectly clean. Perplexity briefly mentions an unrelated company, aINSIGHT, a background-screening service, before correcting itself in the same response. We're including that because it's the real result, not because it needs excusing, an AI system self-correcting mid-answer is a different thing entirely from confidently stating the wrong company by name, which is where this started.
Frequently Asked Questions
Did the fixes cause Google to start getting it right?
Both changed in the same week: eleven specific fixes went live, and Google's answer corrected itself. We're telling you exactly what we did and exactly what happened next, in order, rather than only stating the outcome.
Why did you skip LocalBusiness schema?
Because it's meant for businesses with a physical location customers visit, and we're not one. Adding it would have been technically wrong, not thorough.
Can someone else run this same test?
Yes. Ask the same question across a few AI tools, more than once, before and after you make changes, and record every answer, including the ones that come back wrong or empty.
Why publish the part where Google returned nothing at all?
Because it's the actual starting point, and it's the same starting point most new sites are at. Showing exactly what changed from there, and what it took, is more useful than skipping to the ending.
This is one company, one site, one test. But it points at a bigger question worth asking about search itself right now. Is SEO dead or still relevant in 2026
Most tools in this space will tell you whether AI knows about your business. We built Aensight to go further: find exactly what is stopping it, fix it, and check again. On our own site, that meant eleven specific fixes, applied in one week. Every AI system that mattered was describing us correctly within days of that, and it still is.