How To Use AI In EdTech Without Losing Trust
Most EdTech companies treat their AI usage like a trade secret. They quietly use Large Language Models to generate content, avoid mentioning it publicly, and hope no one asks too many questions. The instinct is understandable; there is a stigma around AI-generated educational content, and for good reason. But secrecy is the wrong response to a legitimate concern.
Some teams have started taking the opposite approach: publishing their full editorial processes, including exactly how and where AI is used, on dedicated pages anyone can read. Platforms have made their entire editorial standards publicly available. This should be normal. Here is a framework any EdTech team can adopt to use AI responsibly and transparently.
In this article...
- The Problem With Hiding AI Use In EdTech
- A 4-Step Editorial Process For AI-Assisted Content In EdTech
- Why Every EdTech Team Needs A No-Fabricated-Citations Policy
- Why Publishing Editorial Standards Works
- A Framework For Other EdTech Teams
- The Bigger Picture
The Problem With Hiding AI Use In EdTech
The concern about AI in educational content is straightforward: Large Language Models hallucinate. They produce text that sounds authoritative but may be factually wrong. They fabricate citations. They present contested claims as settled fact. In education, where the entire purpose is to convey accurate information, these are not minor issues.
But the solution is not to avoid AI entirely. AI is genuinely useful for content creation; it accelerates drafting, helps structure complex topics, and lets small teams produce material at a pace that would otherwise require much larger headcounts. The solution is to use AI responsibly, and to be honest about it.
When EdTech companies hide their AI usage, they create two problems. First, they lose the opportunity to demonstrate that they have safeguards in place. Second, they erode trust with the learners and educators who eventually discover that the content was AI-assisted. And they always discover it because AI-generated content, when not properly reviewed, has tells. Odd phrasing. Overconfident tone on nuanced topics. Citations that do not exist.
Publishing editorial standards is, in part, a trust-building exercise. But more than that, it is a forcing function. When a team commits publicly to a specific process, they actually have to follow it.
A 4-Step Editorial Process For AI-Assisted Content In EdTech
A robust editorial process for AI-assisted educational content typically takes two to four hours per piece. Here is what each step involves.
Step 1: Topic Research
Before anything is drafted, the team should identify the subject, define the scope, and gather primary and secondary sources. For a piece on a historical event, that means official records, contemporaneous accounts, and reputable scholarship (not a quick skim of Wikipedia). Primary sources should be prioritized. Key claims should be cross-referenced against at least two independent sources. For specialized topics, expert review is essential. This step should be entirely human. AI should not select topics or evaluate sources.
Step 2: AI-Assisted Drafting
This is where AI enters the workflow. Large Language Models can help structure and draft content based on the research gathered in Step 1. AI helps move from a collection of notes and sources to a coherent narrative structure more quickly than writing from scratch.
Critically, the AI should never be treated as a source of factual information. It is a writing tool, not a research tool. The distinction matters. Teams should not ask AI "What happened during the California Gold Rush?" and publish the answer. Instead, they should feed it verified information and ask it to organize that information into a readable format.
Step 3: Manual Fact-Checking
Every claim in the draft must be verified against reliable sources. This is the step that separates responsible AI-assisted content from irresponsible AI-generated content.
Teams should check dates, names, and statistics against authoritative references. Quotations should be verified against original texts. Scientific claims should be validated against peer-reviewed research. Reviewers should look for logical consistency and flag anything that feels overconfident or oversimplified. When the AI has introduced errors, they must be corrected.
This step catches more issues than most people expect. AI models are particularly prone to subtle errors: getting a date wrong by a year, attributing a quote to the wrong person, conflating two similar but distinct events. These are exactly the kinds of mistakes that undermine educational credibility, and they are exactly the kinds of mistakes that slip through without human review.
Step 4: Editorial Review
The final step is a full editorial review for clarity, tone, and readability. Does the piece teach what it claims to teach? Is the narrative engaging? Is it pitched at the right level for the target audience? Would a curious adult finishing this piece feel like they genuinely learned something?
A senior team member should own this final review. Direct involvement from leadership in content quality sends a signal, internally and externally, that accuracy is not negotiable.
Why Every EdTech Team Needs A No-Fabricated-Citations Policy
One commitment deserves special attention because it addresses what may be the single most dangerous behaviour of AI in educational contexts: fabricating citations. Large Language Models routinely generate references that do not exist. They will cite books that were never written, attribute findings to studies that were never conducted, and reference journal articles with plausible-sounding titles that are entirely fictional. In an educational context, it is a serious integrity failure.
Teams should commit to never publishing AI-generated references or citations without verifying that the source exists and supports the claim made. Every citation should be checked by a human before publication. This sounds like it should be obvious, but in practice, it is rare. Many platforms that use AI to generate educational content do not have a comparable policy, or if they do, they do not publish it.
Why Publishing Editorial Standards Works
Teams that have adopted public editorial standards consistently report several benefits.
- It raises the bar.
When a process is public, cutting corners feels different. There is no internal conversation about whether to skip fact-checking on a "simple" topic. The published standards become the minimum, and everyone involved in content creation knows it. - It builds trust with audiences.
Learners who care about accuracy respond to transparency. In a market crowded with AI-generated content of questionable quality, a visible editorial process is a genuine differentiator. - It starts conversations.
When teams publish their processes, other founders and content creators reach out to discuss their own approaches. The more companies that publish their processes, the better the industry becomes at holding itself accountable. - It forces honest self-assessment.
No process is perfect. AI-assisted drafting introduces risks that purely human writing does not. Publishing standards creates accountability; when errors are found, they should be corrected publicly, with timestamps, as a visible corrections policy requires. That accountability is uncomfortable but necessary.
A Framework For Other EdTech Teams
If you are building educational content with AI assistance, here is what the evidence supports:
- Publish your process.
Not a vague statement that you "use AI responsibly," but a specific, detailed description of where AI is used, where humans intervene, and what safeguards are in place. Vague assurances are worth nothing. Specific commitments can be evaluated and held to account. - Separate drafting from fact-checking.
AI is good at the former and unreliable at the latter. Treat them as distinct steps with distinct standards. Never let the speed of AI drafting compress the time allocated to human verification. - Verify every citation.
This is nonnegotiable for educational content. If a citation was surfaced or generated by AI, confirm that the source exists and that it supports the claim. If you cannot verify it, remove it. - Have a corrections policy.
You will make mistakes. How you handle them matters more than whether you make them. Commit to a specific timeframe for corrections, mark corrections visibly, and give your audience a way to report errors. - Let humans have the final word.
AI should accelerate your process, not replace your judgment. The moment you remove human oversight from educational content creation is the moment you stop being an education company and start being a content mill.
The Bigger Picture
The EdTech industry is at a crossroads. AI makes it possible to produce educational content at unprecedented scale and speed. That is genuinely exciting. But scale without quality control is not a contribution to education; it is a contribution to misinformation.
The companies that will earn long-term trust are the ones willing to show their work. Not just the polished output, but the process behind it: where AI helps, where humans intervene, and what standards govern the whole operation. If publishing frameworks like this one encourages even a few more teams to open up their processes, the educational content ecosystem will be better for it.
