The drafting is the easy half. Since 1 January 2026 an Ontario posting has mandatory contents, a banned requirement, a response deadline and a retention period — and using a third party to advertise does not move the obligation.
Key takeaways
A model will produce a competent job description from a twenty-minute intake call, and that is a real saving for an agency writing several a week. What it will not do is know which jurisdiction the role sits in or what that jurisdiction now requires. In Ontario, the Employment Standards Act requirements for publicly advertised job postings set mandatory contents, ban a common requirement, impose a response deadline and a three-year retention period — and generally apply to postings made on or after 1 January 2026.
So the workflow is: model drafts, a checklist enforces the statutory content, and a named person confirms the jurisdiction before anything is posted.
After the interview, the employer must tell an interviewed applicant whether a hiring decision has been made for that posting, within 45 days after the interview or, where there is more than one, after the last interview — and where interviews are more than 45 days apart, after each one. It may be given in person, in writing or using technology, including a status update on a portal so long as the applicant has the link.
Records: a copy of every publicly advertised job posting and any associated application form must be kept for 3 years after public access is removed, including each revised version and any linked source referred to in the posting; the information provided to interviewed applicants must be kept for 3 years from the day it was provided.
This is the first Ontario employment rule that speaks directly to a screening tool, and it is drafted around use, not around who owns the software. If an agency ranks, filters or scores applicants on a client’s behalf, the posting has to say so. Two practical steps follow. Ask the question at intake — will anything automated touch this shortlist — and record the answer against the posting. And keep a plain-language description of what the tool actually does, because “we use AI” in a posting invites a question you should be able to answer.
Note what the rule does not do: it does not authorise automated rejection. A model ranks and summarises; a person decides who is interviewed and who is hired, and can explain why.
A generated description reproduces the language of its training material, which includes a great deal of job advertising that would not survive scrutiny here. “Young, energetic team”, “native English speaker”, “must be able to lift” where lifting is incidental, and Canadian-experience phrasing all appear unprompted. The OHRC policy on the Canadian experience barrier explains why the last of those is a discrimination issue and not only an ESA one, and Treadstone Law summarises the protected grounds that a posting can offend.
Accessibility is a further layer: Treadstone Law sets out the AODA employment standards that apply when hiring, including notifying applicants about the availability of accommodation. And if the role is being framed as contract rather than employment, read how classification is actually decided before the posting says so.
A Toronto recruiting agency runs eight to ten searches a month for clients ranging from twelve to four hundred employees.
Intake calls are recorded and transcribed. A model turns the transcript into a structured description — purpose, responsibilities, requirements that are genuinely requirements, and what the first ninety days look like — and flags every requirement it could not tie to something the client said. The compliance fields are not written by the model: compensation, the AI-screening answer, the vacancy statement and the headcount come from a form the client completes, and the posting cannot be published until they are filled.
The agency uses a shortlisting tool, so every Ontario posting for a client above the threshold carries the disclosure. Interview dates create a 45-day task automatically. One client under twenty-five employees is out of scope for the ESA requirements, and the file records why. Nothing about the human rights screen is automated away — a consultant reads the final text.
Recruitment workflow generally is covered under recruitment automation, and internal people processes under HR automation. This page is about the document itself and the rules that govern it. Name the province: everything quoted here is Ontario, and an employment lawyer in the relevant jurisdiction should review a template before it is used at volume.
Does the AI-disclosure rule apply to us or to our client?
The obligation in the ESA guidance is the employer’s, and it says expressly that where an employer uses a third party or a third-party website, the employer is responsible for ensuring the requirements are followed. In practice that means the agency has to give the employer the information to comply.
Do these rules apply outside Ontario?
No. The Employment Standards Act is Ontario legislation. Other provinces and the federal jurisdiction have their own regimes, and a template written for Ontario should not be posted elsewhere without checking.
Can we say “Canadian experience preferred”?
Not in an Ontario publicly advertised job posting or its application form. The OHRC policy explains the human rights analysis that applied even before the ESA rule.
A 30-minute call is enough to tell you whether AI pays for itself here.
