AI-Written Job Ads and Candidate Messages: Where It Helps and Where It Backfires

Recruitment agencies are using AI to write job adverts and outreach messages at scale. Some of that is genuinely useful. Some of it is quietly damaging client relationships, putting wrong information in front of candidates, and making experienced consultants sound like a chatbot. This article covers both sides honestly.

Where AI earns its place in recruitment copy

Turning a vague client brief into a readable first draft

Most job briefs arrive incomplete. A hiring manager sends over a two-line email, a copy-pasted internal job description from three years ago, or a list of bullet points that reads more like a wish list than a real role. A consultant still has to post something by end of day.

AI is genuinely good at taking that raw material and producing a structured draft. It will impose a logical order: role summary, responsibilities, requirements, what the company offers. It will fill in a working format so the consultant has something to edit, not a blank page to fill. That is a real time saving, and it is where the tool earns its place.

The important word is draft. The draft is a starting point, not a finished advert.

Rewriting a long or jargon-heavy spec

Internal job descriptions are written for HR systems, not for candidates. They are full of grading language, internal codes, and corporate phrasing that means nothing to someone reading a job board at 7 in the evening. AI can strip that out and rewrite it in plain language quickly.

Give the model the original spec and ask it to rewrite for an external audience, removing internal references, shortening sentences, and making responsibilities concrete. The output is usually a solid second draft. A person still needs to read it and check it against what the client actually said, but the mechanical work of simplifying dense text is done.

Producing multiple versions for different channels

A LinkedIn advert, a job board post, and a speculative outreach email for passive candidates all want different lengths and different angles. Writing three from scratch takes time. AI can produce all three from a single brief, which a consultant then reviews and adjusts. That is a legitimate use of the tool.

First-draft outreach at volume

If a consultant is running a high-volume campaign for a common role, AI can draft the first version of a candidate message that the consultant then personalises before sending. The draft handles structure. The consultant adds the specific line that makes the message feel individual.

Where it visibly backfires

Identical outreach at scale

The problem with AI-written candidate messages is not that they are AI-written. The problem is that they sound identical when you send hundreds of them without editing them.

Experienced candidates, especially passive ones who are not actively looking, can spot a templated message within two sentences. The structure is the same. The phrasing is the same. "I came across your profile and thought you might be a great fit" appears so often in inboxes now that it has become noise. The candidates most worth reaching are the ones most likely to archive it unread.

AI copy written for volume and sent at volume without human adjustment does not scale outreach. It scales rejection.

Invented role detail the client never confirmed

This is the most serious practical failure, and it happens more often than consultants realise.

AI models generate plausible-sounding detail. When the brief is thin, the model fills gaps with reasonable-sounding content: salary ranges, benefits, team sizes, reporting lines, growth opportunities. None of that may be accurate. The client never said it. The model inferred it from similar roles it has seen.

An advert that promises a management pathway the client has not mentioned, a team of a size the client has not confirmed, or a salary range the client has not agreed to is an advert that creates a misrepresentation problem before you have spoken to a single candidate. It also damages your relationship with the client when they read it.

Requirements that drift from the brief

Related to the above. AI rewrites can quietly shift the requirements. A "preferred" qualification becomes "essential". A "nice to have" becomes a listed requirement. A three-year experience threshold becomes five. The model is pattern-matching against what similar jobs usually say, not faithfully reflecting what this client said.

If a consultant sends that advert without checking it line by line against the original brief, they are posting requirements the client did not set.

Tone that reads as automated to exactly the candidates you most want

Senior candidates, specialist candidates, and passive candidates all have well-tuned filters for automated content. AI copy has recognisable patterns: a specific kind of enthusiasm, a tendency toward abstract phrases like "dynamic environment" and "forward-thinking team", a structure that feels assembled rather than written. It reads as polished in a way that experienced readers find impersonal.

The candidates who are easiest to reach with templated copy are often the ones least hard to find. The candidates worth the most to your clients are the ones who will notice when a message does not feel like it came from a person.

The check that catches invented detail before a client sees it

Run a simple source check before any AI-drafted advert leaves your desk.

Read the advert. For every specific claim, ask: where did this come from? Is it in the brief, the job description, or the notes from the client call? If you cannot point to a source, remove it or mark it as a question to put back to the client.

Create a short checklist your team runs on every AI draft before it goes to the client for sign-off:

• Salary range: confirmed by client, not inferred.
• Benefits: listed only if the client mentioned them.
• Team size and structure: confirmed, or removed.
• Reporting line: confirmed, or described in general terms that commit to nothing.
• Requirements: matched exactly to what the client said, including whether each is essential or preferred.
• Any claim about culture, growth, or opportunity: sourced from the client, not generated by the model.

That check takes five minutes. It is the difference between a draft that builds trust with your client and one that creates a correction conversation you did not need.

A worked example

Here is a generated advert for an account manager role, and what happened when a consultant made three specific changes.

Generated version

"We are looking for a dynamic Account Manager to join our forward-thinking team. You will manage a portfolio of key clients, driving revenue growth and delivering exceptional customer experiences. The ideal candidate will have a proven track record in account management, strong communication skills, and a passion for building relationships. We offer a competitive salary and excellent opportunities for career progression in a fast-paced environment."

This says nothing. Every phrase applies to every account manager role at every company. It makes claims ("excellent opportunities for career progression") that the client never confirmed. It will not help a strong candidate decide whether to apply, because there is nothing to decide on.

After three specific changes

Change one: replace every abstract phrase with a specific fact from the client brief. The client confirmed: twelve existing clients, average contract value of around 80,000 per year, the consultant is the main contact for renewals and upsell conversations.

Change two: state what the role is actually hard. The client mentioned that the accounts have been underserviced and some relationships need rebuilding. That is real context a good candidate wants to know.

Change three: remove any claim the client did not make. Career progression was not mentioned. Cut it.

"Account Manager, [City]. You will take over a portfolio of twelve B2B clients. Average contract value is around 80,000 per year. Your main job is renewals and identifying upsell opportunities. Several of these relationships have not been actively managed for a while, so part of the role is rebuilding trust with clients who are used to being underserviced. You will report to the commercial director and work without a lot of day-to-day supervision. The client is looking for someone who has done this before and is comfortable working autonomously."

That is a different advert. It is honest about the challenge. It gives a senior candidate something to assess. It makes no promises the client did not make. It will attract fewer applications, and better ones.

Using AI in recruitment copy without the downsides

AI is a drafting tool. It is not a research tool and it is not a fact source. It is very good at structure and very bad at accuracy when the brief is thin.

The agencies that get the most out of it are the ones that treat every AI output as a draft requiring a human check, not a finished product requiring minor polish.

If you want a framework for how your agency should use AI tools across the business, including a usage policy you can issue to your team today, the Plainstart AI Policy for Recruitment Agencies covers the key areas in plain language. It is free.

A note on the guidance in this article

This article is general business guidance. It is not legal, professional, or compliance advice. Privacy rules, advertising standards, and employment law vary by jurisdiction, and some of what your agency can and cannot say in a job advert will depend on rules specific to where you operate. Check with an adviser who knows your situation.

AI adoption, done properly.