06-reference/research

teachers pay teachers ai policy scope effective date

2026-11-07·research-brief·source: deep-research·by Ray Data Co (deep-research synthesis)
scribble-workscompetitor-scanmarketplaceai-policyprintables

Thirty days after announcing it, TPT has published no AI content policy at all: no clause, no scope, no effective date, and the only AI text it has published in the window is a training-data promise pointing the other way

The question

"What does Teachers Pay Teachers' forthcoming AI ban actually cover, and does it have an effective date?"

Open follow-up carried forward from [[2026-10-31-marketplace-free-curated-vs-paid-generated-line]], which found the ban existed only in trade press. This run is the delta since then, re-checked against TPT's own published text.

What we already know (from the vault)

What the web says

Convergences and contradictions

Synthesis for RDCO

The clean answer to the question asked is: nothing, and no. TPT's forthcoming AI ban covers nothing in published policy terms, because thirty days after the Oct 8 town hall there is no AI clause in the Seller Guidelines, the Content Guidelines, the Community Guidelines, or a Terms of Service still stamped 12/08/2025, and exactly one of 419 help-center articles mentions AI at all. There is no effective date, and the only TPT-sourced statement about timing is a deferral. That is a complete answer to the backlog item, and the useful move now is to stop treating the ban as a known rule in our own positioning, because it is not one.

The more interesting finding is what TPT did publish. Its single AI help article, created 2026-10-29 and updated 2026-11-04, is a promise not to train competing models on sellers' resources. In the exact window where a content ban was expected, the publishing energy went to reassuring sellers about extraction rather than to defining the prohibition. Read alongside the enforcement language from the town hall transcript, "targeting the lowest quality resources" and "not going to carelessly take action against proven stores," the shape that emerges is not a provenance ban at all. It looks like a quality-signal enforcement regime announced in provenance language, which is consistent with IXL's pre-existing practice of using "algorithmic tools to identify and demote stores associated with low-quality, AI-generated content." If that is what ships, the operative test is detectable badness, not whether a model touched the file.

That distinction is the one thing here with real consequence for Scribble Works, and it cuts in our favour. A provenance-gated marketplace norm would be existential for an AI-generation product in this category; a quality-gated one is survivable, and ruling 14's critic requirement is already the mechanism that answers it. The parent brief landed on parent-trust rather than capability as the durable framing; this run adds evidence for why that is right, and it is sharper than "trust sells better." The largest incumbent in the category has spent a month unable to write down what AI-assisted even means, publicly oscillating between "AI-free," "generative AI should not be used to create images or text," and "the lowest quality resources." The definitional problem is genuinely hard, and whoever can point at a verification step rather than a provenance claim does not have to solve it. Our claim is checkable where theirs is not: the critic gate is a fact about what the parent receives, not an assertion about what tools we avoided.

Two cautions on using any of this. First, do not build positioning copy that quotes TPT policy, because there is none to quote, and the one thing worse than ignoring the ban would be characterizing it from trade press. The Tech Edvocate already published that the ban took effect 2026-10-15, which TPT's own unchanged pages refute. Second, the absence verified here is real but perishable: TPT said more would come "in the next few weeks," and that window is now open. The right re-check trigger is mechanical and cheap rather than calendar-driven, since the help-center article count for AI moving off one is a single API call and is a far better signal than another round of trade press.

Why this is in the vault

Closes the fifth open follow-up of [[2026-10-31-marketplace-free-curated-vs-paid-generated-line]] with a verified negative, so the Scribble Works competitive read stops carrying a reported TPT ban as if it were a published rule. It also supplies the evidence for preferring the critic-gate/verification framing over an AI-provenance framing in printables-product positioning, since the incumbent's own three incompatible scope statements show the provenance line is not defensible even by the party drawing it.

Open follow-ups

Related

Sources

Vault

TPT primary sources (all read in-browser 2026-11-07, approx 01:20 ET)

Secondary (trade press and seller blogs, explicitly not policy)

Blocked, flagged, not retried