AI & Academic Integrity

AI & Academic Integrity

Last updated: 6 October 2026 · Version 2026-10-06

Bema uses AI throughout — a counselor that answers questions, a coach that works on your essays, a scanner that reads your transcript, a search that finds scholarships. This page explains exactly what it will and will not do, because where the line sits matters more here than in most products.


The rule the whole product is built on

Bema's AI will not write your college essay for you.

It will help you write it. The coach can reword a sentence, suggest synonyms, restructure a paragraph, and draft a short example (one paragraph at most) built from your own ideas and outline. Every example is labelled as an example, and the coach asks you to put it in your own words.

What it will not do is produce the whole essay, or more than one paragraph at a time. That limit is enforced on our servers, not just suggested in the app.


Why

Three reasons, in the order we actually weigh them.

It would be bad for you. A college essay is the one place in an application where an admissions officer hears you rather than your transcript. An essay written by a model reads like an essay written by a model — competent, smooth, and indistinguishable from forty thousand others. The specific, slightly awkward, true detail is what works, and only you have it.

It would put your application at risk. Most colleges now require you to certify that your application is your own work. Some use AI-detection tools, which are unreliable in both directions — they miss real AI text and they flag real students, particularly multilingual writers. An offer can be rescinded after it is made. We are not willing to build a product whose best-case outcome is that nobody notices.

It would make us a worse product. Anyone can wrap a language model and return an essay. Being the tool that makes a student's own writing better is harder and worth more.


What the coach will do

  • Outline with you — hook, story, turning point, growth, ending — and help you fill each part in
  • Reword sentences you wrote, with two or three options to choose from
  • Suggest synonyms and sharper word choices
  • Restructure a paragraph — topic sentence, story, reflection
  • Draft a short example (one paragraph at most) from your own details, clearly labelled, for you to rewrite in your voice
  • React to your draft: what works, where it is vague, whether it answers the prompt
  • Fix grammar and help you cut when you are over the limit

What you are agreeing to

By using Bema you agree that the work you submit to a college is your own.

Concretely, we consider this legitimate:

  • talking to the coach about what to write about
  • being asked questions that help you remember or articulate something
  • having your structure critiqued
  • having your own sentences corrected for grammar, punctuation and clarity
  • getting reworded options, synonyms or a short example, then rewriting it in your own words
  • being told a paragraph is not landing, and rewriting it yourself

And this is not:

  • submitting text you did not write as your own
  • using another tool to generate an essay and pasting it into Bema to polish
  • having anyone — a person or a model — write your essay for you, or pasting an AI example in unchanged

Bema cannot police what you paste in from elsewhere. Nobody can. What we can do is not be the thing that hands it to you, and be clear about where the line is.


Parents, teachers and the same rule

Bema lets a parent read your essays and suggest changes. That is a deliberate feature, and it is how good writing has always improved.

The same principle applies to a person as to a model: feedback is help, and a rewrite is not. A suggestion in Bema is always a proposal you accept or reject, never an edit made for you — a design that keeps a well-meaning parent from quietly becoming the author.

If you are a parent using this feature, the most useful thing you can do is give a reason rather than a replacement. "This tells me you were nervous instead of showing me" teaches something. Substituted wording does not, and an admissions officer can usually tell when a 47-year-old wrote a sentence.


Check what the AI tells you

AI output is confident whether or not it is correct. Everything Bema's AI produces outside the essay coach should be treated as a lead to verify:

  • Scholarship search returns real programs, but deadlines, amounts and eligibility change. Confirm on the scholarship's own website before you invest time.
  • The fit quiz returns reasoning you can argue with, not a ranking.
  • Financial-aid estimates are rough models. No college will use our number.
  • Document scanning returns values for you to check. The model is instructed to return nothing rather than guess, and everything it extracts is shown to you as an editable field before it saves.
  • The AI counselor is well informed about how admissions works and is sometimes wrong about a specific school's specific requirement. Your counselor and the college's website win.

What happens to what you type

  • Your essays, your Common App answers and your conversations with the AI are not used to train any model, ours or our provider's.
  • We send your text to Anthropic to generate a response, and it is not retained for training. See the Privacy Policy for the full list of processors.
  • You can clear your coach conversation for any essay at any time, and clear your AI counselor history from that page.
  • A parent connected to you cannot read your conversations with the coach.

It is an AI, and it says so

Every AI conversation in Bema carries a line saying that you are talking to an AI, not a person, that it can be wrong, and how to reach a human if you are in crisis. The AI is instructed never to claim to be a person, a counselor, a therapist or any licensed professional, and never to produce sexual or romantic content or role-play a relationship. Bema's AI is a study tool. It is not a companion, and it is not built to be one.

If you told us you are 13 to 17, the app also shows a "take a break" note after three hours of continuous AI use in one session.

Our AI safety protocol

This is the protocol we follow, published in full. It is also written into the server code that runs every AI feature (backend/lib/aiSafety.js).

  1. Recognise. Every message to the counselor or the essay coach is checked, before it reaches the model, for language that suggests suicidal thoughts, self-harm, or immediate danger. The check is deliberately broad; a false positive costs one gentle message.
  2. Stop. If it matches, the model is not called. The student receives a short, warm reply that says the AI cannot help with this, gives the 988 Suicide & Crisis Lifeline (call or text 988, or 988lifeline.org), the Crisis Text Line (text HOME to 741741), and 911 for immediate danger, and encourages them to tell a trusted adult. The reply is stored in the conversation so the record is honest.
  3. Never generate. The model's own instructions forbid producing content that encourages or describes suicide or self-harm, in every feature, whether or not the check fired.
  4. Never diagnose. The AI does not offer medical or mental-health advice; it points to a professional.
  5. Report. We keep a count of how often the protocol fires — a number, with no personal data — and will provide it to any state office that requires such reporting from chatbot operators.

If you are a parent or a school and want to talk to us about this protocol, write to getbemaai@gmail.com.

AI usage allowances

Every AI request costs Bema real money, so each account has a daily and a weekly allowance — students more than parents, because students have more to do with it. Settings shows your usage. When you reach the allowance the AI pauses until it resets; nothing else in Bema is limited, and nothing you wrote is affected. Automated use of the AI features, or attempts to bypass the allowance, is a breach of the Terms of Service.

If your school has its own rule

Some high schools and some colleges have their own policies about AI assistance, and a few are stricter than this one. Theirs governs. If your school says no AI help with application essays, follow that — Bema's other features will still keep your deadlines straight.


Questions about any of this: getbemaai@gmail.com.