Platforms Say 'Don't Optimize'? Smart Teams Are Quietly Running Experiments

Hi everyone, this is Neo.

Lately, a louder and louder voice in the SEO world is telling us: “Don’t bother optimizing for AI search. Skip the whole ‘content chunking’ thing. Just write good content that’s useful to humans, and AI will naturally recommend you.”

Sounds nice, right?

But as a veteran who’s spent years in the independent-site and traffic game, I have to tell you: listen, but don’t take it too seriously.

Why? Because platforms sit on the side of “ecosystem stability,” while you sit on the side of “survival and growth.” Those two interests don’t always align.

Today I’m breaking down a very hardcore article that cites a fresh research paper (E-GEO) and reveals a brutal truth: while platforms tell you “don’t optimize,” smart teams are already running scientific experiments — and getting stunning results.

01 Why “Content Is King” Is a Correct but Useless Statement?

We all know “content is king,” but in the AI search era, the meaning has changed.

Before, you’d write a sprawling 10,000-word article, Google would rank it on page one, and users would click in and read slowly. That was the “page ranking” game.

Now, a user asks AI: “Which running shoes should I buy?” AI hands back a ready-made summary. The user’s journey might end right on the search results page. This is the “answer selection” game.

No matter how great your content is, if AI can’t read it, can’t extract the key points, or can’t stitch it into a “best answer,” you’re out of the game entirely.

That’s why we can’t just follow platforms’ “zen” advice. We need to take the initiative and take control of GEO (Generative Engine Optimization).

02 The E-GEO Experiment: Using Data to Punch Holes in the Mysticism

The article cites a paper called E-GEO: A Testbed for Generative Engine Optimization in E-Commerce.

Instead of just theorizing, this paper did something very concrete:

  1. It gathered a bunch of e-commerce product pages.
  2. Rewrote the content of those pages using different strategies.
  3. Measured which rewrite made AI more willing to recommend the product.

The results were fascinating:

  • Many classic “rewrite” tricks (like padding in some filler or shuffling word order) completely failed in front of AI — some even backfired.
  • But when pages were optimized with a “structured decision support” template, the improvement was striking.

The research found that AI doesn’t like flashy marketing language (like “ultra-premium” or “top-tier experience”). It prefers clear, specific content with constraints and comparisons.

Neo’s take: This shows AI isn’t looking for “well-written” articles — it’s looking for “usable data.” It needs to feed your content to users to help them decide. If your content is full of vague adjectives, AI will struggle to extract useful information. But if your content is as logically tight as an instruction manual, AI will see you as “reliable” and cite you first.

03 The Million-Dollar “Universal LLM Optimization Template” (8 Steps)

This is the best part of the article. Based on the research, the author distilled a “universal LLM rewrite recipe.” It’s not just a writing template — it’s a standard for turning your product into “AI-friendly” data.

Bookmark this and use it to optimize your core product pages:

Step 1: Define the purpose in one sentence (Purpose)

  • ❌ Wrong: “This is a coffee machine with an ultimate experience.”
  • ✅ Right: “This fully automatic espresso machine is designed for small households, perfect for users who need a quick cup every morning.”
  • Key point: be explicit about the scenario, constraints, and core value.

Step 2: List your selection criteria (Selection Criteria)

  • Write it like a “spec sheet” — in plain language, list the metrics users care about most.
  • If it’s a knife, talk about steel hardness and edge retention; if it’s software, talk about integrations and learning curve.

Step 3: Say the “bad stuff” up front (Constraints)

  • ❌ Wrong: hiding the downsides at the end.
  • ✅ Right: “This product is not suitable for high-frequency commercial use,” “in low-light conditions it performs worse than a professional DSLR.”
  • Key point: machines hate ambiguity. Explicitly telling AI who your product is not for actually increases the odds it gets recommended to the right people.

Step 4: State what it is and what it isn’t (Definition)

  • Template: “This is designed for [audience A], not for [audience B].”
  • This massively reduces ambiguity.

Step 5: Turn selling points into testable claims (Testable Claims)

  • Don’t say “durable” — say “tested through 100,000 presses.”
  • Don’t say “fast” — say “cold start in just 3 seconds.”

Step 6: Provide comparison hooks (Comparison Hooks)

  • Help AI compare proactively: “Compared to [competitor A], our advantage is [X], but we fall slightly short on [Y].”
  • This gives AI ready-made material to cite.

Step 7: Evidence anchors (Evidence Anchors)

  • List certifications, warranty terms, and concrete specs. This is how you build trust.

Step 8: Decision shortcut (Decision Shortcut)

  • Template: “If you care most about [A and B], choose us; if you care more about [C], consider [other type].”
  • Directly help the user (and AI) make the final call.

Neo’s take: See? This template completely flips traditional “sales copy.” It’s no longer about moving people with emotion — it’s about arming AI with logic and facts. Really, it’s about saving AI effort: once you’ve organized the answer for it, AI will naturally serve your answer to users.

04 Don’t Just Read About It — Test It!

The article ends with a very sensible suggestion: don’t blindly trust this template — test it.

Your industry, your brand authority, and your audience are all different, so results will naturally differ.

Neo’s recommended action plan:

  1. Pick 10-20 similar product or article pages.
  2. Split them into two groups. Leave one group untouched; rewrite the other using the “8-step method” above.
  3. Watch for a month and compare how the two groups perform in AI search (Perplexity, ChatGPT Search, Google AI Overviews, etc.).
  4. Focus on: is your content being cited by AI? Is it being presented as an answer?

Summary

The winds have changed. SEO gets you “indexed” (seen), while GEO gets you “selected” (recommended).

Don’t be fooled by platforms’ “don’t optimize” smoke screen. The real upside always belongs to the few who dare to run experiments at the edge of the rules and find the patterns.

I’m Neo — in this era of AI reshaping traffic, let’s evolve together.

References:

  • Search Engine Journal: When Platforms Say ‘Don’t Optimize,’ Smart Teams Run Experiments
  • arXiv: E-GEO: A Testbed for Generative Engine Optimization in E-Commerce