Blooncyclopedia:Generative AI

The Anti-Bloon is better for popping Bloons than writing strategy sections.

To preserve the integrity of Blooncyclopedia's writing standards and file standards, we generally do not allow users to add AI-generated or AI-assisted content to the wiki. This is mainly because text generated by LLMs is often strangely written, outdated, and sometimes straight-up incorrect; citing AI-generated content (such as articles and strategy guides on other websites) is also ill-advised, for similar reasons. Regarding AI-generated images, they have absolutely no place on a wiki that focuses on assets and screenshots that are minimally edited and as accurate to the source material as possible.

Most LLMs are trained on material that is not released under a compatible license, leaving the copyright status of content generated by these models unclear. For example, asking an LLM to explain the mechanics of an upgrade in BTD6 will likely result in text derived from content scraped from the Bloons Wiki on Fandom, which is incompatible with this wiki's license. Thus, LLM-generated text is considered plagiarism for the purposes of this wiki unless it was generated by a model that was trained exclusively on content available under a compatible license.

Note that these guidelines do not apply to documenting the rare cases where AI-generated content appears in official Bloons media. However, avoid speculating about what official content is AI-generated; any claims that some official material is AI-generated requires explicit confirmation or strong evidence supporting it.

AI-generated content in pages

On the whole, you are not allowed to add AI-generated and AI-assisted content to any pages. This includes, but is not limited to:

• Text generated by an LLM
• AI-generated or upscaled images
• Machine translation, such as machine translating official material that isn't in English (such as the Chinese version of Bloons TD 6)

Using AI in discussion of the wiki itself, on talk pages and forums, is also forbidden. Copyright issues aside, generating responses to other people's arguments with ChatGPT is always a bad look. Even if you are not fluent in English, explaining your perspective in your own words always comes across as more sincere than having an AI do it for you.

Acceptable uses of AI

While you may use AI as the basis for the following, please review the content, correct any errors, and make sure it matches the formatting of the rest of the page before adding it to the wiki. If something looks so obviously AI-generated, it will probably be removed.

• Formatting content into wikitext, such as tables and galleries
• Checking for spelling or grammatical errors
• Lua code

For example, if you are not fluent in English, you may use AI to detect and fix errors in text that you already wrote in English, but you should not write something in your native language first and then machine translate it into English.

Identifying AI-generated text

In general, do not assume something is AI-generated just because it is written strangely. Always ask others for their opinion before removing any text that you think is AI-generated. However, if you do see common signs of AI writing, it's a good idea to start a discussion, either on the village pump or the talk page of the page in question or the user who added it. Examples of common signs:

• Some models, especially ones integrated into search engines like Google Gemini, introduce a topic with a sentence like "here's a breakdown of X" or "here's why X" followed by a list of bullet points, usually with the topics of each bullet point written in bold.
• ChatGPT in particular overuses em dashes, transition words, parallelism (such as the sentence structure "It's not just X, but Y"), flowery words like "unpack" or "delve", and exclamation points.
• Overuse of bullet point lists. Bullet point lists are often useful for organizing lots of small bits of disjointed information, but LLMs often do this even when writing things like summaries that would be better off written in full paragraphs.

Keep in mind that some common signs of AI writing, such as inconsistent tone and switching between second and third person, could simply be due to the wiki's nature as a collaborative effort; since any user can add to or change sentences, these "red flags" may just be a consequence of the text being written by multiple writers writing different parts. Always check the article's revision history and editors' contributions before making any accusations that their work is AI-generated.

See Wikipedia:Signs of AI writing for more information on identifying AI-generated text.