Wikipedia:Yes, you have to follow NOLLM

Essay on editing Wikipedia From Wikipedia, the free encyclopedia

A very common response when users utilising large language models in editing are confronted with the WP:NOLLM guideline is some variation of the following:

Thank you for your concern. However, please understand that WP:NOLLM is a guideline, not a policy, and it outlines best practices rather than rules that must be followed without exceptions. After using an LLM to write an article, I verify the sources manually and edit the text in accordance with WP:V, WP:NPOV and WP:MOS. I believe that common sense should be applied in this case and what I have inserted is genuinely verifiable, encyclopaedic text, not unverifiable hallucinations. If you have any specific concerns regarding the article itself, please feel free to let me know.

One way of dealing with this sort of user is to painstakingly find the style and verifiability issues present in the article itself, which will almost certainly be present. This is, however, time- and energy-consuming, in a disproportionate manner compared to the effort the editor has put in to generating the article text. A more straightforward and obvious approach is to call out the user for being disingenuous, because they have admitted to violating NOLLM already.

WP:PAG states:

Although Wikipedia generally does not have hard-and-fast rules, policies and guidelines are standards all users should normally follow, with guidelines providing guidance in specific contexts. Policies and guidelines should be applied using reason and common sense.

The section on adherence also specifically states:

Use common sense in interpreting and applying policies and guidelines; rules have occasional exceptions. However, those who violate the spirit of a rule may be reprimanded or sanctioned even if they do not technically break the rule.

Therefore, WP:NOLLM is not a "best practice"; it is a rule in the exact same way that not promoting fringe theories, conflict of interest, assume good faith, and even reliable sources are rules. No one, no matter how new or experienced or technically competent, may use an LLM to directly generate article text. Insistence on doing so may lead to sanctions.

Citing this essay

This essay applies to the case where the editor has admitted to NOLLM-violating behaviour but is sealioning or otherwise refusing to follow the guideline. It does not apply to the case where LLM use is simply suspected. False positives happen, and assuming good faith is just as much of a guideline as NOLLM is. However, in the case where the editor has admitted to violating and continuing to violate a guideline, one should not be afraid to call a spade a spade, or even call for sanctions where disruption is ongoing.

If this essay is cited on you

As mentioned above, NOLLM applies to you. You are not an exception, no matter how competently you think you are using LLMs, and the evidence on the AI noticeboard is that people consistently think they are much more LLM-competent than they are. You must stop using LLMs to generate article content, or face sanctions. On the off chance you are consulting an LLM on matters of Wikipedia policy, this should serve as a lesson that LLMs don't know anything about Wikipedia policy.

See also what to do if you are accused of using an LLM.

The necessity of this essay

To some extent, this essay states the obvious. However, cursory experience at the AI noticeboard will tell you that these obvious facts are not at all obvious to many LLM users, up to and including admins. This essay provides a quick and centralised refutation to such arguments.

See also

Related Articles

Wikiwand AI