Legal

How AI review works

The AI pre-submission review reads a project's current drawings and reports before a submission to the City and flags likely revisions. A flag is a likely revision, not an approval, a legal opinion or a professional sign-off; the City decides, and your professionals remain responsible for the submission.

What it checks

• Completeness: the documents the City's Application Checklist requires for site plan control, given the project's size and use.
• Zoning: numeric standards from Zoning By-law 569-2013 (parking, loading, bicycle parking, and the CR zone's density and height) against the project's site statistics and the City's open zoning data.
• Discipline checks: questions about the drawings and reports, answered by an AI model against the rule texts, with the Toronto Green Standard's mandatory items among them.

Every rule comes from a published source, with its section and the version date, and every flag cites the rule it relies on and, where the drawings show it, the sheet. Rules we couldn't verify are never used. Things the City can't require at site plan stage are never flagged.

Reading the results

• Blocking: something the City requires is missing, or a numeric standard isn't met. Expect a revision.
• Likely revision: a rule the drawings or reports seem not to meet. Check it before you submit.
• Advisory: from a City guideline rather than a by-law or a requirement, so it's never marked blocking.
• Can't check yet: the review couldn't tell, and names what it needs, such as a site statistic or a sheet.

Any flag can become a requested change with an owner and a due date, or be dismissed with a reason. When you import the City's comment letter, ApproveIQ compares the City's comments with the review's flags, so you can see how well it predicted them.

What it costs

A review is paid in credits. The button shows the price before you run it, worked out from the number of sheets and reports and the checks that apply. If part of a review fails, the credits for that part come back to you automatically.

What it can't check

• Scanned sheets without a text layer (only page-level flags).
• Lots under a former zoning by-law, and the text of site-specific exceptions: the review says so instead of guessing.
• Guidelines: anything from a guideline is advisory, never blocking.

Where documents go

Stored files stay in Canada. For a review, the documents it reads are sent to Anthropic's Claude API for processing in the United States; the review screen asks for your agreement before the first run. Anthropic's handling of API data is described in its API data retention documentation. Customer documents are never sent through Anthropic's batch or file storage services.

Learning from past approvals

The review is designed to learn from past approved submissions by retrieving how the City commented on similar projects, using only material we have the right to use. It doesn't train a model on your projects.