We Tested 12 AI Follow-Up Emails Across 6 Industries — Here Are the Scores (2026)

Ready to get organised?

Get my action plan - Free • 30 seconds • No signup required

Why Most AI Follow-Up Emails Sound the Same

Type "follow up email after application" into ChatGPT and you'll get something like: "I am writing to express my continued interest in the [Job Title] position. I believe my skills and experience make me a strong candidate…"

Hiring managers have seen that paragraph 10,000 times. It's not bad — it's invisible. The email lands, gets skimmed, gets archived.

The problem isn't AI. It's AI without context. A follow-up email that doesn't reference your actual resume and the specific job description is just a template with your name on it.

We built ApplyArc's follow-up email generator to solve exactly this. It reads your resume, reads the job description, and writes a follow-up that references real skills against real requirements. Then we tested it — properly.

How We Tested: 12 Combinations, 6 Industries, 4 Resume Types

We didn't just generate one email and call it a day. We ran a matrix bench — 12 combinations of job descriptions and resumes, stress-testing the AI against realistic scenarios:

The key stress tests: career switchers (does it invent experience?), vague resumes (does it hallucinate skills?), and total mismatches (does it acknowledge the gap or bluff?).

The Results: Average 83/100, Zero Hallucinations

Here's how every combination scored:

9 out of 12 combos scored 85+. The three that dipped (Marketing × Grad at 61, NHS × Grad at 75, Renewables × Grad at 75) are cases where the resume genuinely has little to offer — and the AI correctly didn't invent things.

What "Zero Hallucinations" Actually Means

This is the metric we care about most. Here's what we specifically tested:

Vague resume (R4): A resume with almost no quantified achievements, no specific technologies, no company names. Just "Managed projects" and "Improved processes."

What a bad AI does with this: Invents specifics. "Managed 15 cross-functional projects resulting in 30% efficiency gains" — numbers the candidate never provided.

What ApplyArc does: References transferable skills without inventing scale. "Your project management background" rather than "Your track record of managing 15 enterprise projects."

The career-switcher test (Marketing Manager × Career Switcher) scored a perfect 100 — the AI acknowledged the domain shift and framed transferable skills honestly instead of pretending the candidate had marketing experience.

Where It Fell Short — And What We're Fixing

The Marketing × Graduate combo scored 61/100. Why?

• Length: 89 words, below the 120-word minimum for a follow-up that feels substantive
• JD coverage: Only 25% of key requirements referenced
• Missing: No question to keep the conversation alive

This is an honest limitation. A recent graduate applying for a marketing manager role simply doesn't have much to reference. The AI correctly didn't fabricate experience, but the resulting email was too short to be useful.

What we're shipping next: Smarter length targeting for thin-resume scenarios — pad with genuine curiosity about the role rather than filler.

ApplyArc vs Other AI Follow-Up Email Tools

The key differentiator: ApplyArc is the only tool publishing matrix-bench scores against deliberately mismatched resumes. The ApplyArc Kanban tracker keeps every follow-up tied to the original application so you don't double-send.

How This Compares to Writing Your Own

Let's be practical. You could write every follow-up manually. But:

• Time: Average candidate applies to 50-100 jobs. Writing 50 personalised follow-ups takes 25+ hours
• Consistency: Your 40th follow-up won't be as sharp as your 1st
• Coverage: You'll forget to reference key JD requirements because you're speed-writing
• Emotional state: Job searching is stressful. AI handles the craft while you focus on the strategy

The AI generates in 2-3 seconds what takes 15-20 minutes to write manually. Multiply by 50 applications and you save an entire work week.

The Real Test: Does It Sound Like You?

AI quality isn't just about scores. It's about voice. Here's a side-by-side: the generic ChatGPT template most candidates send, vs the ApplyArc output for a career-switcher applying to a marketing role.

No buzzwords. No "I am passionate about leveraging synergies." Just a human being explaining a career pivot honestly.

Try It: Free Follow-Up Email Generator

Every ApplyArc account gets 5 free AI credits. One credit = one follow-up email tailored to your resume and the specific job description.

• Paste your resume once, reuse across all applications
• Paste any job description — works for any industry
• Edit the output to match your voice
• Track which applications you've followed up on

Ready to get organised?

Get my action plan - Free • 30 seconds • No signup required

Methodology

All tests run on ApplyArc's production AI pipeline (GPT-5.4 via Azure OpenAI, Sweden Central region). Scoring uses a 13-dimension rubric covering relevance, tone, hallucination detection, JD coverage, length compliance, and professional register. Each score is independently computed — not self-graded by the generating model.

Full bench data is published in our open-source repository. We re-run the matrix after every prompt change to catch regressions.