A lot of AEO and GEO work underperforms for a simple reason: it’s aimed at the wrong problem. Teams treat AI search as if it’s one channel with one set of ranking inputs and one user journey. It isn’t.
What people call “AI search” is a bundle of different systems. They surface answers in different ways, pull from different sources, and reward different types of evidence. They also change with every new model, and the experience varies by paid tier and audience. If you try to force one playbook across all of that, you end up executing a generic strategy that gets mediocre results (or none at all).
How this goes wrong in AEO and GEO
In practice, it looks like reasonable work applied in the wrong place. Common examples include:
- Rewriting a handful of pages to be “more conversational” and expecting visibility lifts across every AI surface.
- Copying a competitor’s FAQ pattern without checking whether the category is actually winnable, and what “winning” would mean commercially.
- Assuming a single schema template will cover every product line, service variation, and intent type.
- Measuring success with one metric (for example, impressions or referrals) while the business outcome is leads, calls, or qualified demos.
None of these tactics is inherently wrong. The mistake is treating them as universal.
No two AI environments reward the same evidence
If you want AEO and GEO to perform, you need to start from how each environment decides what is “safe” to surface. Some systems lean heavily on sources they already trust. Others prefer content that is easy to extract and summarise. Others still behave more like traditional search, but with a different presentation layer.
The durable takeaway is this: you are not optimising for a single algorithm. You are building a body of evidence that can be reused across multiple answer systems.
If you have limited resources (like most of us do), pick the highest value channel for your audience and just focus there. Rather than doing an average job chasing results on every platform.
Expect different experiences by tier and audience
Even within the same platform, what gets shown can vary depending on paid tier, geography, device, and whether the AI is personalising the response using ‘memories’ as a result of past history. That matters because it changes what “winning” looks like.
If your reporting assumes one universal SERP-like experience, you will misread the results. You might be improving visibility in one context while losing ground in the one that drives revenue.
Category competitiveness still decides how hard this is
Under new interfaces, the old variable still does most of the work: how competitive your category is. Two businesses can perform the “same” AEO tasks and achieve completely different outcomes because their baselines differ.
In a low-competition niche, basic clarity and structured content can move the needle quickly. In a high-competition category, you need stronger proof, tighter differentiation, and better coverage of the decision journey. The work becomes less about wordsmithing and more about building a credible, referenceable footprint.
Same category does not mean the same rules
Even inside one category, the rules change by sub-intent. “What is it” queries behave differently from “which one should I buy” queries. “How much does it cost” behaves differently from “is it compliant” or “does it integrate with X”.
If you don’t segment content by intent, you’ll over-invest in content that gets surfaced but doesn’t convert, or you’ll ignore the intent that actually converts the deal.
Content authority over keywords
Often there’s no clean primary keyword to target at all. That’s increasingly common with AI-driven discovery, because demand is fragmented across a long string of unique phrases and follow-up questions.
When you don’t have a clean primary keyword, the goal shifts from “rank for X” to “be the best source for a set of decisions.”
High competition = More complexity
We see this constantly: teams apply a generic “AI optimisation” checklist and get noise instead of results. When we take a category-specific approach, performance becomes measurable.
As one example, we worked with a nationwide manufacturer on a custom plan that combined new content, site-wide schema, and rebuilt FAQs aligned to real buyer questions. Over three months, their ChatGPT referral sessions increased by 354%. It wasn’t because of one trick. It was because the work matched the category, mapped to intent, and the way answer engines retrieve evidence.
What to do this week if your AEO plan feels stuck
If that sounds like your current plan, don’t start by producing more content. Start by checking whether you’re solving the right problem.
Pick one high-value product or service line and do three things: identify the 10–15 questions that show up in real sales calls, rewrite or build one page that answers those questions with specifics, and then connect it to supporting pages with consistent language and structured data where appropriate. Give it a few weeks, measure the business signals, and iterate.
That approach is slower than copying a template. It’s also the difference between “we did AEO” and “AEO increased our pipeline”.
