What recruiters actually notice in AI job applications
Yes, most hiring teams can spot AI-written applications, but not because of some secret detection software. They can tell because the writing sounds like everyone else who pasted a prompt into the same chatbot. The tool is not the problem. The lazy output is.
What screening tools actually detect
There is a whole industry built around "AI detection" software, and most of it is unreliable. In our experience, these tools flag genuine human writing as often as they catch actual AI output. They measure things like sentence length variation, vocabulary distribution, and something called "perplexity," which is a rough proxy for how predictable the word choices are. The problem is that predictable writing is not the same as AI writing. A tired applicant writing their tenth cover letter at midnight produces low-perplexity text too.
Applicant tracking systems, the software most employers use to collect and filter applications, do not currently run AI detection as a standard filter. As of 2026, the major ATS platforms focus on keyword matching, resume parsing, and structured data extraction. They care whether your resume contains the words from the job description. They do not care whether a human or a machine arranged those words.
What does get flagged is not AI itself but plagiarism patterns. If your cover letter contains the exact same phrases as fifty other applicants who used the same prompt template, that gets noticed. Not by AI detection software, but by the recruiter who has read forty cover letters that day and recognizes the sentence "I was particularly drawn to this opportunity because of your company's commitment to innovation and excellence."
Why generic AI prose gets filtered
AI writing has tells. They are not subtle once you have read enough of them. Here is what recruiters and hiring managers actually notice:
- Template openings. "I am writing to express my interest in..." or "I was thrilled to see..." or the worst one, "In today's rapidly evolving [industry] landscape." These are not wrong, they are just everywhere.
- Abstract enthusiasm. "I am passionate about leveraging synergies to drive meaningful impact." This says nothing. A human who actually wants the job mentions something specific about the role or company.
- Symmetric structure. Every paragraph is roughly the same length. Every bullet point follows the same grammatical pattern. Real writing is messier than that.
- Hollow achievements. "Successfully collaborated with cross-functional teams to deliver key initiatives." What teams? What initiatives? What did you actually do?
- Overqualification theater. The AI does not know what you do not know, so it confidently describes experience you do not have. If you let it write unchecked, it will invent a leadership background for someone who has been an individual contributor for two years.
The filtering happens not because a computer caught you, but because a human skimmed your application, felt nothing, and moved on. Generic writing is forgettable writing. In a stack of two hundred applications, forgettable is the same as rejected.
What humans actually notice
Hiring managers are not running your text through analysis tools. They are reading with their eyes and their gut. Here is what actually tips them off, in order of how obvious it is:
First, the voice does not match the interview. If your cover letter reads like a polished corporate press release and you stumble through a phone screen saying "um" every three words, the gap is jarring. This is the most common way AI-assisted applications fall apart. The writing sets a bar the candidate cannot clear in conversation.
Second, the details are wrong or missing. AI is good at structure and bad at specifics. It will write you a beautiful paragraph about your "extensive experience in project management" without knowing that your actual project management experience was three months of herding volunteers for a community garden. A human writes about the garden. The AI writes about "stakeholder alignment."
Third, everyone sounds the same. If a recruiter reads ten applications in a row and eight of them have the same rhythm, the same transition phrases, the same bullet point structure, the two that sound different stand out immediately. Not necessarily because they are better, but because they are recognizable as individual people.
How to use AI as a drafting tool without sounding like everyone else
AI is genuinely useful for getting started. The problem is not using it, it is using it wrong. Here is a workflow that works, based on what we have seen across thousands of applications:
Give it your raw material, not a prompt
Do not ask the AI to "write a cover letter for a marketing role." Instead, dump your actual notes into the prompt. "Here are three things I did at my last job: I ran a campaign that increased email open rates by fixing our segmentation, I built a spreadsheet to track which blog posts converted, and I argued with my boss for six months about switching to a different email tool and eventually won. Here is the job description. Help me organize these into a cover letter."
The output will still need editing, but it will be grounded in your actual experience instead of invented generalities.
Edit aggressively, especially the first and last paragraphs
AI tools reliably produce the worst openings and closings. The opening will be some variation of "I am excited to apply for..." and the closing will be "I look forward to discussing how my skills can contribute to..." Rewrite both. Start with something specific. End with something concrete.
A good opening might be: "Your job posting mentions you need someone to clean up a messy CRM migration. That is what I spent the last year doing at my current job, and it went poorly enough that I learned a lot about what not to do."
That is not polished. It is human. It tells the hiring manager something specific. It sounds like a person.
Cut the adjectives and abstractions
AI loves adjectives. "Comprehensive," "strategic," "innovative," "dynamic," "results-driven." Go through the draft and delete every adjective that is not earning its keep. If you cannot replace "strategic" with a description of the actual strategy, cut it.
The same goes for abstract nouns. "Collaboration," "initiative," "impact," "excellence." These words mean nothing without specifics attached. "Led a team of four" beats "demonstrated leadership capabilities" every time.
Read it out loud
This is the cheapest and most effective AI detection tool available. If you read your application out loud and it sounds like a corporate press release, it needs rewriting. If you would not say the words in a conversation with a real human, do not put them in your application.
The goal is not to sound unprofessional. It is to sound like a professional who is also a person. Those are different things.
Where automation actually helps
Using AI to draft applications is not the same as automating your entire job search. Drafting assistance is low risk. You review the output, you edit it, you send it yourself. The problems start when automation removes you from the loop entirely.
At Reach, we use automation to handle the parts of job searching that are genuinely mechanical: finding postings that match your criteria, filling in repetitive form fields, tracking where you have applied and when. The applications themselves still go through human review because, in our experience, the moment you stop reading your own applications is the moment they start sounding like everyone else's.
The line is not between AI and human. It is between "AI helped me produce this" and "AI produced this and I hit send." The first is fine. The second is how you end up submitting a cover letter that says "As a passionate advocate for synergistic stakeholder engagement" to a company that makes garden hoses.
Practical checklist
Before you submit anything, run it through this list:
- Does the first sentence say something specific? If it could appear in any cover letter for any job, rewrite it.
- Are there real numbers, names, or details? Not invented ones. Actual specifics from your experience.
- Did you delete the adjectives? Most of them, anyway. Keep only the ones that add information.
- Would you say this out loud? In an actual conversation with a hiring manager, not in a speech.
- Does it sound different from the last five applications you read? If you cannot tell, read five more and compare.
If your application passes these checks, it does not matter whether AI helped you write it. What matters is that a human can read it and recognize another human on the other end. That is the whole point.