I Stopped Optimizing for Rankings. Here’s What I Optimize for Now.
By Daniel Haiem, CEO, AppMakers USA
For most of last year, I watched our content rank well and still lose visibility. Pages held position one. Traffic kept sliding. It took me longer than I’d like to admit to understand what was actually happening: the search engine wasn’t sending fewer people to find the answer. AI Overviews, Perplexity, and ChatGPT were finding the answer and never sending anyone to find it at all.
That distinction changed how I think about content entirely, and it’s the reason most SEO advice right now, including some AI SEO strategies, still feels like it’s solving last year’s problem.
Ranking and Being Cited Are Not the Same Goal
For two decades, ranking was the only scoreboard that mattered. Get to position one, and the traffic followed. That logic is breaking down quietly, page by page, because AI systems don’t browse a ranked list and click through. They read content, extract the answer, and synthesize a response. Your page can be the single best source on a topic and still generate zero clicks, because the AI system already gave the user what they needed.
This isn’t a ranking problem. It’s a structural problem. Content built to rank is built around keyword density, internal linking, and on-page signals that search engines reward. Content built to be cited has to be built around something different: extractability. Can an AI system isolate a clean, self-contained, accurate answer from your page without needing the surrounding context?
Most content can’t do that, because nobody wrote it to.
What Actually Gets Pulled Into AI Answers
When I started auditing which of our pages were showing up in AI Overviews and which weren’t, the pattern wasn’t subtle. The content that got cited had a few things in common that the content getting ignored did not.
It answered the question in the first two sentences, not the fifth paragraph. AI systems extract early, and content that buries its answer under three paragraphs of preamble rarely gets pulled, even if the preamble is well-written and the answer is correct.
It used specific numbers, not vague claims. “Improved significantly” doesn’t get cited. “Reduced page load time by 40%” does. AI systems favor content with concrete, attributable specifics because specifics are easier to extract cleanly and easier to trust.
It was structured around a single, well-defined question per section. Content that tries to cover five related ideas in one flowing paragraph is harder for an extraction system to isolate cleanly than content that breaks each idea into its own clearly headed section.
None of this is exotic. It’s closer to how a good encyclopedia entry is written than how a good blog post is written. That’s the shift most of us haven’t fully made yet.
Topical Authority Now Means Something More Specific
“Build topical authority” has been SEO advice for years, but in an AI search context, it means something narrower than publishing a lot of related content. AI systems appear to weight content more heavily when it comes from a source that has demonstrated consistent, first-hand expertise on a specific topic over time, not just broad coverage of a category.
In practice, that means a site with twelve generic articles about software development gets less credit than a site with four articles that go deep on a specific, narrow problem the author has clearly lived through. Depth and demonstrated experience are outperforming breadth, and that’s a real change in how content strategy should be prioritized.
I’ve started telling clients to cut their content calendars by half and double the specificity of what remains. Fewer, sharper, more experience-grounded pieces are outperforming a higher volume of competent-but-generic ones.
The Metric I Actually Watch Now
Traffic dashboards still show rankings clearly. They show almost nothing about citation visibility, which is the metric that increasingly matters. I’ve started tracking a rougher proxy instead: how often our brand name, our specific claims, or our exact phrasing shows up when I query ChatGPT, Perplexity, and Gemini directly on questions our content addresses.
It’s manual. It’s not as clean as a rank tracker. But it’s the closest thing I have right now to measuring whether our content is actually doing its job in the environment where a growing share of searches are happening.
What I’d Tell Anyone Starting This Audit Today
Pick your five highest-value pages. Ask whether each one answers its core question in the first two sentences. Check whether your claims are backed by specific numbers an AI system could extract cleanly. And ask honestly whether the page reflects something you’ve actually done, not just something you’ve researched and summarized.
The content that wins in this environment isn’t the content that’s most optimized for traditional search. It’s the content that’s most extractable, most clearly grounded in real experience, and best positioned for AI SEO visibility.
About the Author
Daniel Haiem is the CEO of AppMakers USA, a mobile app development agency that works with founders on mobile and web builds. He is known for pairing product clarity with delivery discipline, helping teams make smart scope calls and ship what matters.
Related Reading:
Want to understand how AI search works and what businesses can do to improve visibility in ChatGPT, Gemini, and AI Overviews? Read: What Is AI SEO? A Beginner’s Guide to Ranking in AI Search
