Canada's federal research granting agencies recently did something unusual for a government funder: they dropped a rule. Their earlier guidance had asked applicants to declare whether they used generative AI in preparing a proposal. The updated version removed that blanket requirement, with the agencies noting that mandatory disclosure could cause uncertainty, inconsistency and unnecessary administrative burden. So, can you use AI to write a grant in Canada? For most funders, yes, but the rules that remain are stricter than they look, and they all point to the same thing: you own every word you submit.
This article summarizes what Canadian funders have actually published about AI in grant applications, where the real risks are, and a practical list of do's and don'ts for nonprofits, charities and community groups.
What Canadian Funders Say About AI Grant Writing
Most community foundations and provincial programs have not published a dedicated AI policy yet. Two national funders have, and their guidance is a good indicator of where others are heading.
The federal research agencies (CIHR, NSERC, SSHRC and CFI)
The guidance on the use of artificial intelligence in the development and review of research grant proposals, shared by the Canadian Institutes of Health Research, the Natural Sciences and Engineering Research Council, the Social Sciences and Humanities Research Council and the Canada Foundation for Innovation, makes a few points that apply well beyond academic research:
- AI is allowed for applicants. The agencies recognize that generative AI can improve efficiency and help applicants writing in their second language.
- The named applicant is accountable for everything. Applicants must make sure the information in their application is true, accurate and complete, and that sources are properly acknowledged.
- No blanket disclosure requirement, with one exception. If an AI tool surfaces specific works or sources that significantly inform your proposal, you are expected to find and reference those original sources.
- Protect sensitive information. Applicants are warned to check the privacy and data-retention practices of online tools before entering confidential material.
- Reviewers may not use public AI tools. Uploading an application into a publicly available AI tool to evaluate it is prohibited because it breaches confidentiality.
The Canada Council for the Arts
In December 2025 the Canada Council for the Arts published its own guidance on AI in grant applications, built around principles of fairness, accountability, security and transparency. It accepts that applicants may use AI tools for support tasks such as outlining, grammar, plain-language editing and translation, and that this can reduce barriers to applying. It also sets clear limits: applicants remain responsible for the accuracy and originality of their applications, should not enter confidential information into public tools, and must not present AI-generated content as solely their own original work without attribution. Where AI is part of the funded project itself, applicants are expected to describe its role and address rights and ethical questions.
No Canadian funder we have found bans AI outright. Every one of them puts the full weight of accuracy, originality and confidentiality back on the applicant.
Everyone else
Many foundations, municipalities and provincial programs are silent on AI. Silence is not permission to submit anything. Their existing rules about accuracy, attestation and conflict of interest still apply, and many application forms ask a signing officer to certify that the information is true and complete. If a funder has not published a policy and you are unsure, ask the program officer. Always confirm current guidelines on the funder's website before applying.
Where AI Actually Helps
Used well, AI tools are useful for the parts of grant writing that are mechanical or repetitive:
- Unpacking guidelines. Turning a 30-page program guide into a checklist of eligibility rules, required attachments and scoring criteria, which you then verify against the original.
- Outlining. Mapping each question to the evidence you have, before you start drafting.
- Trimming to character limits. Online portals often cap answers tightly. AI is good at cutting a 2,400-character answer to 1,500 without losing the point.
- Plain-language editing. Removing jargon and making sentences easier for a volunteer reviewer to read quickly.
- Translation support. A first pass for a French or English version, reviewed by a fluent human.
- Consistency checks. Asking whether the numbers in your narrative match your budget, or whether your outcomes match your activities.
Where AI Gets Organizations Into Trouble
Invented facts and statistics
AI tools can produce plausible but false information, sometimes called hallucinations. A made-up statistic about youth participation in your region, or a citation to a report that does not exist, can sink an application and damage your credibility with that funder for years. Every number needs a real, checkable source. Our guide on demonstrating community need with data funders trust explains where to find Canadian data that holds up.
Generic writing that sounds like everyone else
Reviewers read dozens of applications in a sitting. As more applicants lean on the same tools, the same phrases repeat: "underserved communities", "holistic approach", "meaningful impact". What separates a funded application is specificity: your participants, your town, your waitlist, your coach who drives the van. AI does not know any of that unless you tell it, and even then it tends to sand off the details that make your case memorable. See how funders actually score applications for what reviewers reward.
Privacy and confidentiality
Pasting participant stories, health information, financial statements or a partner's confidential budget into a public AI tool can expose information you are obligated to protect. Both the federal agencies and the Canada Council flag this risk. If you would not email it to a stranger, do not paste it into a public tool.
Letters of support written by the applicant
Using AI to draft letters of support "from" partners and asking them to sign is a fast way to produce five letters that read identically. Funders notice. Give partners a short brief and let them write in their own voice. Our article on letters of support covers how to make that easy for them.
A Practical Checklist: Do's and Don'ts
| Do | Don't |
|---|---|
| Read the funder's AI guidance, if it has any, before you start | Assume silence means anything goes |
| Use AI to outline, trim and edit your own draft | Submit a generated draft you have not rewritten |
| Verify every statistic against the original source | Cite a number or report you have not opened yourself |
| Remove personal and confidential information first | Paste participant stories or partner budgets into public tools |
| Add local detail only you know | Let generic phrasing replace your real story |
| Be transparent when a funder asks, or when AI produced a substantial part | Present generated text as entirely your own original work |
A Simple Workflow That Keeps You Safe
- Gather your raw material first. Program data, participant numbers, budget, staff bios, partner commitments and two or three real stories (with consent).
- Write the core answers in your own words, even if roughly. This is where your voice and specifics live.
- Use AI to tighten, not to originate: shorten to limits, fix grammar, flag unclear sentences.
- Fact-check line by line. Every number, name, date and program rule against the source.
- Have a human read it cold. Someone outside the project should be able to follow it and believe it.
- Record what you did. A short internal note on which tools were used and how makes it easy to answer a funder's question later.
Frequently Asked Questions
Is it allowed to use AI to write a grant application in Canada?
Generally yes. The federal research agencies and the Canada Council for the Arts both accept that applicants may use AI tools, but the applicant remains fully responsible for accuracy, originality, proper sourcing and protecting confidential information. Check each funder's own guidance.
Do I have to disclose that I used AI?
It depends on the funder. The federal research agencies removed their blanket disclosure requirement but expect you to reference original sources that an AI tool surfaced. The Canada Council says you must not present AI-generated content as solely your own original work without attribution. When in doubt, ask the program officer.
Can funders tell if an application was written by AI?
Reviewers may not run a detector, but generic phrasing, missing local detail and unverifiable statistics are easy to spot and score poorly regardless of how they were produced.
Can grant reviewers use AI to score my application?
Not with public tools at the federal research agencies, which prohibit reviewers from uploading applications to publicly available AI tools. The Canada Council has separate guidance for its assessors.
AI can save you time on the mechanical parts of an application, but the story, the evidence and the accountability still have to be yours. If you want a team that writes from your real programs and checks every fact, Book a 10-minute discovery call.