Technical SEO Has Never Been More Important: Here’s Why
While everyone is hunting for signals of what works in the AI era using tools of questionable reliability, others are looking in familiar places to uncover insights about the new players reshaping search.
Technical SEO is currently one of the few remaining bastions of reliable AEO data. Here’s what to look at:
Prompt/Query Tracking Under Pressure
One of the biggest developments of the past two years has been measuring citation and mention rates inside AI assistants.
Various tools regularly fire the prompts you give them into ChatGPT, Gemini, Claude, and the like, to see whether your site or brand appears. You show up: success.
The criticism, however, has been relentless, with many accusing this visibility metric of being yet another “vanity indicator”, delivering no real value, much like organic traffic once was.
A recent Sparktoro study, published just a few days before I wrote this (February 1, 2026), piles on further, concluding that AI systems rarely return the same list of brands and recommendations twice.
So what do you do if you believe prompt tracking in a tool like Profound is little more than noise? The answer: retreat to your logs.
The Best Insights Come From Log Files
Log analysis: feared by some for its complexity, revered by those who actually understand it, offers unparalleled detail on how LLMs interact with your site.
Jérôme Salomon of Oncrawl, himself skeptical of prompt tracking, goes even further, arguing that the practice pollutes log files, which he considers the only truly reliable way to understand AI behavior (see original post).
We now know that OpenAI’s three bots have a significant impact on what ChatGPT users actually see:
- OAI-SearchBot – responsible for citations
- GPTBot – focused on content for model training
- ChatGPT-User – handling custom GPTs, but especially web searches triggered by users
Imagine making changes to your key pages and then noticing a higher crawl rate shortly after. A little while later, your conversions from ChatGPT climb by X%…
Your logs will tell you which bots are crawling, when, and help you build theories far closer to reality than anything most other practices, simpler, perhaps, but far more debatable, could ever produce.
It’s Not That Daunting
Log analysis and the technical work that comes with it can feel intimidating to the uninitiated, but it’s not enormously difficult. The steepest part of the curve is usually wrestling with tools that have terrible interfaces — but that’s just exploration, and then habit.
As for the upfront investment: depending on your site’s size, the cost can sting (anywhere from $100 to $20,000+). But if your employer, or you, if you run the business, is still hesitating and struggling to justify the expense, the clock is ticking. Every major player in the space is already doing this and uncovering valuable insights every month.
The longer you wait, the longer you delay understanding what actually works for your business — while your competitors are almost certainly already there.
The path forward is straightforward:
- Subscribe to a log analysis tool (even a free one)
- Export your logs, sanitize them (i.e., strip out sensitive data), then load them into your tool — ideally with automated daily ingestion — to start getting data
- Isolate the performance of OpenAI, Gemini, and Perplexity AI bots
- Run optimizations, then observe crawler behavior
- Keep what works and measure the impact on conversions
Far more concrete, even if some gray areas remain, like cause-and-effect interpretations, which are unavoidable at this stage.
And setting log analysis aside: page speed, Schema markup, and content duplication are all widely suspected to influence LLM decision-making.
So… when do you start?