AI Usage Policy

Last updated: May 2026 · Author: Marcus K., Editor-in-Chief

The Pokémon GO content space has been heavily affected by AI-generated text since 2023. Most “best Pokémon GO” listicles, tier-list rewrites, and “ultimate guide” pages on the open web are now partially or fully generated by language models, then lightly edited and published. The result is a flood of plausible-looking content that is wrong about specific things in specific ways — hallucinated counters, stats from previous metagames, moves that no longer exist, recommendations that contradict themselves between paragraphs.

This page describes Pogodex’s position on AI tools: where we use them, where we do not, and what is disclosed.

What we do not do

We do not generate editorial content with AI. Specifically:

We do not use language models to write guides, tier lists, raid counter lists, Pokédex entries, event coverage, or analytical features. The editorial voice on Pogodex is human, and the analysis behind it comes from in-game testing by our editors.
We do not use AI to “expand” short notes into long-form content. If we have something short to say, we say it short.
We do not use AI to generate “summary” or “TL;DR” sections that the human writer would not have written themselves.
We do not use AI to generate Pokémon stat tables, move lists, or any numerical content. Numerical content comes from data files (verified by Aria) and is not generated.
We do not use AI to generate Pokémon artwork, sprite art, or imagery presented as game content. Where we use imagery, it is either officially permitted promotional artwork (Niantic press kit) or original illustration created in-house.
We do not use AI to write search-engine-targeted “Q&A” sections that the writer does not believe in.

The reason these are absolute commitments rather than soft guidelines: language models are confidently wrong about Pokémon GO in the kind of specific ways that matter to players. They will tell you a Pokémon has access to a move it has never had access to. They will recommend a counter that is below the level cap for the raid being discussed. They will hallucinate Community Day dates. The cost of getting these things wrong, for a reader trying to make a stardust decision or a raid party plan, is real.

Limited operational uses

We do allow narrow, operational uses of AI tools where the failure modes are visible and a human still does the editorial work:

Spell-check and grammar. Standard tooling that has integrated language model features. Used for surface-level cleanup, never for substantive editing.
Translation help for non-English correspondence (we receive reader email in many languages). The editor still reads the original and verifies the translation.
Drafting alt text for our own original images, subject to human review before publication. The image is described accurately or the alt text is rewritten.
Assistance with non-content code (the IV calculator’s underlying math, the stat verification scripts). The output is reviewed and tested before deployment.
Search and exploration: looking for documentation, finding relevant Niantic announcements, locating community discussions of edge cases. The AI is a search assistant, not a content source.

None of these uses generate publishable editorial content. The principle: AI is allowed where a human is checking the output before it reaches the reader.

How AI use is disclosed

Where AI has been used in any way that affects content the reader sees, we disclose:

If we publish a guest contribution that involved AI-assisted drafting, we ask the contributor to disclose it; we publish the disclosure with the article.
If we ever publish a feature where AI was a meaningful research aid (rather than a search tool), we say so in the article.
We do not publish “produced with AI assistance” boilerplate on every article, because it would be untrue for the great majority of our content.

Detection and reader recourse

If you suspect content on Pogodex was generated by AI, please email info [at] pogodex [punto] space with the URL and what triggered the suspicion. We investigate and respond. If you are right, we will say so publicly, retract or rewrite the affected content, and document the failure.

We have no plans to introduce AI-generated content. If our position changes — for example, we decide to publish an experimental section that uses AI for some specific purpose — the change will be announced explicitly, this page will be updated, and the affected content will be marked.

Why this matters for Pokémon GO specifically

Pokémon GO is a game where the difference between good and bad advice is concrete. Bad raid counter advice means lost raid passes. Bad PvP team advice means lost ladder ranking. Bad “is it worth powering up” advice means wasted stardust that takes weeks to grind back. The cost of being wrong is paid by the reader.

The economics of AI-generated Pokémon GO content do not work for the reader. They work for the publisher who wants to flood the search results with plausible pages. We are not interested in that publishing model. We would rather publish less and be right.

Reader-submitted content and AI

Reader-submitted tips, comments, and corrections are accepted at face value but are reviewed before being incorporated into editorial content. If a tip turns out to be AI-generated and incorrect, we discard it and (where possible) note the source. Our comment policy and the broader framework for reader-submitted content are described in our Terms of Service.

If you submit content to us as a guest writer, please be honest about whether AI was involved in drafting it; we do not automatically reject AI-assisted contributions, but we will not present them as fully human-authored if they are not. Hidden AI use in submissions is grounds for declining the contribution and, in egregious cases, for declining future contributions from the same source.