AIO/LLM Simulator Game for Content Teams (Human AI Adoption)
I designed a workshop game to solve a problem that’s costing brands visibility: Google’s AI Overviews are citing Reddit threads and affiliate sites instead of the brand’s own content—even when the brand has better information.
Example: a search for a product category returns a summary implying you can only buy in physical stores, while ignoring the brand’s direct e-commerce offering.
That’s a content + structure problem as much as it is an SEO problem.
So after a colleague’s presentation on AIO/LLM search, I designed a game to make the mechanics stick—built from scratch as a simulation:
- Teams become the AI Overview system
- They get a fixed “token budget” (representing a limited context window—the system can only process so much at once)
- They choose from mixed-quality “source cards” (brand site, Reddit posts, comparison sites, marketing copy, product specs, etc.)
- Their task: write an “AI Overview” where every key claim is tied to evidence
- Then an “evidence drop” forces a revision—because that’s what happens when different sources appear.
Why a game? Because the fastest way I could think of to internalize this is to be the system—to feel the constraints and see how source quality determines what you can confidently say.
The practical takeaways for content teams:
- Write simple, specific facts that can be quoted safely
- Make purchase options explicit (including e-commerce) and hard to miss
- Add structured sections: FAQs, “Where to buy,” constraints, and what to verify before taking action
- Include tradeoffs/edge cases so the model doesn’t fill gaps with generic language (tradeoffs also add credibility)
- Design pages so they can stand alone if they’re the only thing retrieved
If you’re seeing AI Overviews summarize your category from everyone except you, this is solvable—but it requires designing content so AI can use it accurately, not just for human browsing.
I’ve packaged this as a ready-to-run workshop (cards, rubric, facilitator guide). If you want to run it with your team, email me.


(Originally published on LinkedIn, here.)