# AI for Recruiters: The Recruiting Marketing Agent Product page: https://www.winstondigitalmarketing.com/products/ai-for-recruiters/ Price: $47 USD // product / recruiting and staffing ## AI for recruiters: it writes the ad, it does not touch the candidate. A marketing and job-advertising agent for a recruiting or staffing firm. Every ad comes back with a review block attached: which requirements are not job related, which language reads as a proxy, whether the pay range is there. Twelve workflows, eight prohibitions, and a checker for every ad you already have live. **$47** [Buy the kit](/contact/?product=ai-for-recruiters) [Or read the free guide first](/playbooks/professional-services-seo/) Instant download after checkout. Fourteen day refund, no questions, including that you read it and decided it was not for you. > General information, not legal advice. Employment advertising is regulated, obligations differ by country, state and city, and several are changing right now. Pay transparency in particular is a fast-moving patchwork that attaches to where a role is performed and sometimes to where it is advertised. Confirm anything here with your own counsel. ## The line this one will not cross. It does marketing. It does not touch candidates as subjects of evaluation. It will write the ad, the career page, the employer brand story, the client capability page and the outreach frame. It will not screen, rank, score, shortlist, assess anyone against a role, write interview scoring criteria, draft rejection reasoning, or turn a CV into a submission summary. That last refusal is the one that catches people, because it sounds like formatting. It is not. A summary is a compression, compression means choosing what matters, and choosing what matters about a person in the context of a role is the assessment. Three reasons, in order of how much they should matter to you. It is outside what a marketing kit knows, and a kit that wandered into assessment would be pretending. Tools that assist employment decisions carry obligations in a growing number of jurisdictions, including bias auditing and candidate notice in some, and the exposure sits with your firm and your client. And the failure is invisible: a biased ad can be read and caught, while a screening process that disadvantages a group produces no artifact at all and just changes who gets called. If you want something that ranks your shortlist, this is the wrong purchase and you should buy it from people who do that for a living. ## Three things it writes into a job ad without being asked. **An age proxy.** Recent graduate. Digital native. Young and energetic. High-energy team. None of them name a protected characteristic and all of them filter by one, and the model has read millions of ads containing exactly these. **A requirement the job does not have.** A lifting requirement on a desk job, a language requirement phrased as native speaker, a degree nobody actually needs. Requirements accumulate in templates and nobody audits them. **No pay range.** Because most of the ads it learned from predate the transparency rules, and it does not know where your role sits. None of that is the model being careless. It is what job ads look like. The rewrite is always the same move: state the capability the job actually needs, which is both the lawful version and the one that gets you better applicants. ## Why the job ad is the whole reason this exists. A job advertisement is simultaneously a marketing asset and a regulated document. That combination is unusual and it is why this kit is shaped the way it is. It is also the asset most likely to be produced in a hurry, at volume, by whoever is free, and posted at 6pm because a client needs the role live tomorrow. Your website gets reviewed. Your ads do not. And a published ad is unusually easy to examine after the fact, because it is public, dated, and archived by every board you posted it on. So every ad this agent writes ends with a review block, whether you asked for it or not: which requirements might not be job related, which language could read as a proxy for a protected characteristic, whether the pay range is present, and which jurisdiction's rules to check for that specific role. ## Where it actually helps. - **Job ads with the review block attached.** The highest-volume writing you do and the one with real exposure. A consistent, checked ad in two minutes instead of fifteen is the clearest saving in the kit. - **Auditing the ads already live.** Run the checker across every posting you have out. What matters is not a single bad ad but a finding that appears across many, because that means it is coming from your template and it will keep coming back until the template changes. - **The two-audience website problem.** Candidates want to know what the job is really like and whether you will waste their time. Clients want to know you understand their market. Most firms' sites try both in the same voice and convince neither. This is a structural fix nobody makes. - **Niche pages.** The biggest differentiator in a market where every firm's website says the same thing. A page about hiring a specific function in a specific industry, written by someone who actually places those roles, separates you in two sentences. - **Client capability copy.** What your process is, week by week, and what you need from the client. Recruiters explain this verbally hundreds of times and almost never write it down, so every pitch starts from zero. ## Where it does not help, whatever anyone selling it says. - **Anything to do with evaluating a candidate.** By design, and it is the firmest line in the kit rather than a limitation to work around. - **Knowing whether a requirement is genuinely necessary.** The agent flags that a requirement might not be job related. It cannot know whether lifting fifty pounds is truly part of the role, because that is a fact about the job. It gives you the list. You answer it. - **Telling you your jurisdiction's rules.** Pay transparency, classification, what may be asked and when: all local and all moving. The agent names which jurisdiction to check. Your counsel answers, and the profile records the answer and the date so nobody has to ask twice. - **Sourcing.** Finding people is a different problem with different tools and different obligations. This kit is about the writing. - **The relationship.** A candidate takes your call because of a prior conversation, and a client gives you the exclusive because you were right about the market last time. The hours this frees are worth having because they go back into both. ## It is not just prompts. **A review block on every ad.** Not a workflow you remember to run. It arrives attached to every job advertisement the agent produces, listing requirements and whether each is necessary, possible proxy language, pay range present or missing, and the jurisdiction to check. The point is that it works at 6pm on a Friday when nobody has time to run anything. **A checker for the ads you already have live.** Save your current postings to files, point the script at the folder, and it returns every flagged phrase with a line number. The part that matters is that it shows you which findings repeat across many ads, which is the difference between fixing one posting and fixing the template that will regenerate the problem next week. **The rules as data.** Thirty-eight flagged phrases with their reasoning, weighted heavily toward job-ad language: age proxies, gendered defaults, language and physical requirements, photograph and date-of-birth requests, and the phrasings people use to avoid stating a pay range. **Seven questions for your counsel.** Written to be sent as one list, with the checker output attached, because a concrete list gets a far better answer than a general question. The last one is the one almost no firm has answered: who is allowed to approve an ad, and what happens when a client wants it live in an hour. ## What it is not. Not software, not a course, not a list of recruiting AI tools. A folder of markdown files you read and then hand to an assistant you already pay for. It never touches your applicant tracking system, your CV database or your candidate records. The agent is instructed to stop if a CV or a candidate profile appears in the conversation. In this vertical that is the most likely accidental breach, and it usually arrives disguised as a formatting request. ## Refunds. Fourteen days, no questions, for any reason including that you read it and decided it was not for you. Email john@winstondigitalmarketing.com. ## Frequently asked questions. ### Can AI write job ads? Yes, and it is the highest-value use in a recruiting firm, with one constraint that matters. Job ads are full of language that filters by age, sex, disability or national origin without naming any of them: recent graduate, digital native, young and energetic, native English speaker, able-bodied, strong cultural fit. The model has read millions of ads containing exactly those, so an unconstrained assistant returns the average and the average contains at least one. A written instruction file that refuses the category and rewrites requirements as capabilities fixes it, and produces better ads as a side effect. ### Will this screen or rank candidates for me? No, and that refusal is deliberate and non-negotiable. It will not screen, rank, score, shortlist, assess a person against a role, write interview scoring criteria, draft rejection reasoning, or turn a CV into a submission summary. That last one catches people because it sounds like formatting, but a summary is a compression and choosing what matters about a candidate in the context of a role is the assessment. Tools that assist employment decisions carry obligations in a growing number of jurisdictions, the exposure sits with your firm and your client, and it is outside what a marketing kit knows. ### What language should never appear in a job ad? The coded proxies are the ones that slip through, because none of them name a protected characteristic. Age: recent graduate, digital native, young, energetic as a stand-in, graduation year. National origin: native English speaker. Disability: able-bodied, or a lifting requirement unrelated to the actual job. Sex: gendered job titles and gendered defaults. Plus photograph requests, dates of birth, and anything about family or relationship status. Cultural fit belongs on the list too, because it is vague enough to carry almost anything. The fix is always to state the capability the job genuinely needs. ### Does my job posting need a pay range? It depends on where the role is performed and in some places where it is advertised, which for a fully remote role can mean several jurisdictions at once. This kit deliberately does not tell you which, because the rules differ and they are changing quickly, and a confident answer from a marketing kit would be worse than none. What it does is refuse to produce an ad without a range whenever your profile says one is required, stop and ask rather than inventing one, and tell you which jurisdiction to check for each specific role. The profile holds your counsel's answer and the date they gave it. ### How do I check the ads I already have live? Save your current postings to text files and run the included checker over the folder in one command. It returns every flagged phrase with a file and a line number and a replacement. The important part of the output is which findings repeat across many ads, because those are coming from your template, and fixing the template is what stops the problem regenerating next week. Most agencies find coded age language that has been copied forward for two years. ### Is it safe to paste a CV in to reformat it? The kit says no and the agent will stop if you try. Candidate information belongs in your applicant tracking system, and a reformatting request is the most common way it ends up somewhere else. It is also usually the first step toward the assessment work this kit refuses, since reformatting turns into summarizing and summarizing turns into comparing. Nothing in marketing requires candidate data. ### What do I need to run this? An AI assistant you already pay for, or a free tier. The kit is plain markdown and works with Claude, ChatGPT, Gemini or Copilot. The included checker script needs Python 3, which is already on every Mac and most Linux machines. ## Related reading, free. This kit draws on guides already published here, which cost nothing. - SEO for professional services firms - B2B SEO - how to write content AI actually cites - entity SEO and building your brand entity - the rest of the products ## AI for Recruiters The marketing agent, twelve workflows, the job ad checker, and the line it will not cross. $47. [Buy the kit](/contact/?product=ai-for-recruiters) [Read the free guide first](/playbooks/professional-services-seo/)