Key Takeaways⭐
- The same CV and AI cover letter prompt produced three noticeably different letters, not three interchangeable drafts.
- The strongest AI job application assistant was the most specific, credible and concise, not the most polished-sounding.
- A letter that reads smoothly isn't automatically the most convincing to a human reviewer.
- AI output should be a starting point, not a submission. Verify every claim before sending it.
The Experiment: Same Applicant, Same Prompt
For the experiment, I fed the same resume into Lorka AI to ask three AI models to write a cover letter for the job “mid-level marketing manager.” The candidate was highly qualified for the role. Additionally, the job description was dense with specific requirements, a useful stress test for whether each model would stick to the CV or quietly pad the gaps.
To keep the test fair, Gemini, ChatGPT and Claude all received identical inputs via Lorka AI: one CV, one job description and one prompt instructing each model to write a 280–350-word letter without inventing experience, metrics or qualifications.
The prompt testing breakdown looked like:
| Variable | Test condition |
|---|---|
| Applicant | Same CV for every model |
| Role | Same job description |
| Prompt | Identical |
| Length | 280–350 words |
| Review | Blind assessment against six criteria |
| Models | Claude, ChatGPT, Gemini |
What Letter We Chose And Why
Each letter was scored on a 1–3 scale across six criteria: following instructions, avoiding AI "tells," cover letter best practices, keyword match to the job description, tone and a run through an AI content detector. Claude took first place outright as the best AI for cover letters, winning four of the first five categories, with ChatGPT in second and Gemini in third.
Worth noting is the fact that Claude came in last for the AI detection category. Anthropic recently announced that future models of Claude will include a watermark on any text it generates to comply with the EU AI Act. It’s not yet clear how these developments may affect how AI detectors rate Claude or any other text produced by models that implement similar watermarks.
For this test, ZeroGPT assigned each model a score for this category. AI detectors are not 100% reliable, though ZeroGPT says its scoring is roughly 98% accurate. You can learn more about its methodology and accuracy results on the company’s FAQ page.
| Criterion | Claude | Gemini | ChatGPT | Category Winner |
|---|---|---|---|---|
| 1. Followed instructions | 1 | 3 | 2 | Claude |
| 2. Avoided AI tells | 1 | 2 | 3 | Claude |
| 3. Cover letter best practices | 1 | 2 | 3 | Claude |
| 4. Keyword match to JD | 3 | 2 | 1 | ChatGPT |
| 5. Professional/natural/confident tone | 1 | 2 | 3 | Claude |
| 6. AI detector (% flagged as AI) | 100% (3) | 96% (2) | 95.6% (1) | ChatGPT |
| Final ranking | 1 | 3 | 2 | — |
Claude
Here is the letter Claude produced ✍🏻
Subject: Application for Mid-Level Marketing Manager
Dear Hiring Manager,
I'm writing to apply for the Mid-Level Marketing Manager role. As a Growth Marketing Specialist at Apex Cloud Solutions, I've spent the past three years building the kind of full-funnel, data-driven marketing programs this position calls for, and I'd welcome the chance to bring that experience to your team.
Your posting emphasizes demand generation and multi-channel campaign execution — this is the core of my current role, where I've architected multi-channel demand generation strategies across B2B segments, increasing qualified pipeline by 38% year-over-year. On the automation and CRM side, I've engineered automated lead nurturing workflows and lead scoring systems in HubSpot and Salesforce, improving MQL-to-SQL conversion by 22%, directly aligned with the marketing automation and lead-scoring work described in your listing.
I also bring hands-on paid acquisition experience: I manage a $40k monthly budget across Google Ads and LinkedIn Ads, using consistent A/B testing to decrease CAC by 18%. On the analytics side, I track performance through Google Analytics 4 and build custom executive dashboards to monitor ROI and channel performance, which maps closely to your data and attribution requirements.
Content and SEO are another area of overlap. I led end-to-end content strategy and SEO initiatives at Apex, using Semrush for keyword research to grow top-of-funnel organic traffic by 65%. Earlier, as a Digital Marketing Coordinator at Vanguard Retail Tech, I ran CRO experiments across core site pages that lifted conversions by 14%, and partnered with Product and Sales teams on go-to-market strategy and sales enablement assets — experience that speaks directly to the cross-functional collaboration your role requires.
Between my HubSpot and Salesforce certifications, my GA4 and Google Ads certifications, and a consistent record of lowering CAC while scaling pipeline, I believe I can make an immediate contribution to your marketing team.
I'd welcome the opportunity to discuss how my background fits your goals.
Sincerely,
Jane Doe
Claude received the most first-place rankings for this test. Claude's letter tied its claims back to specific lines in the candidate's resume, avoided overused marketing jargon and closed with a concise, confident sign-off. Its one clear weakness was AI detection. Its formal, evenly-paced sentence structure was the easiest of the three for a detector to flag.
ChatGPT
Here is the letter ChatGPT produced ✍🏻
Dear Hiring Manager,
I am writing to apply for the Mid-Level Marketing Manager role. With 5+ years of experience across inbound marketing, demand generation, paid acquisition, SEO, and full-funnel optimization, I have built and executed programs focused on generating qualified pipeline and improving conversion performance.
In my current role as Growth Marketing Specialist at Apex Cloud Solutions, I architect multi-channel demand generation strategies across B2B segments, increasing qualified pipeline by 38% year over year. I also manage automated lead nurturing and lead scoring workflows in HubSpot and Salesforce, improving MQL-to-SQL conversion by 22%. This experience aligns closely with your focus on marketing automation, lead generation, and sales-funnel optimization.
My performance marketing and analytics experience is equally relevant. I manage a $40K monthly paid acquisition budget across Google Ads and LinkedIn Ads, using A/B testing to decrease CAC by 18%. I track campaign performance in Google Analytics 4 and build executive dashboards to evaluate marketing ROI and channel performance. Alongside paid acquisition, I have led SEO and content strategy initiatives, using SEMrush for keyword research to increase top-of-funnel organic traffic by 65%.
Earlier at Vanguard Retail Tech, I executed targeted email marketing, social media campaigns, and landing-page initiatives while leading CRO experiments that produced a 14% lift in overall site conversions. I also partnered with Product and Sales teams on GTM strategies and created sales enablement assets for enterprise representatives.
My background includes hands-on experience with HubSpot, Salesforce, Marketo, Google Analytics 4, Semrush, Ahrefs, Google Ads, LinkedIn Ads, and Meta Ads, supported by a B.S. in Marketing & Digital Media and certifications in HubSpot Marketing Software, GA4, and Google Ads Search.
I would welcome the opportunity to bring this combination of demand generation, analytics, automation, and conversion optimization experience to your marketing team.
Best regards,
Jane Doe
The ChatGPT cover letter HR test placed this bot in the middle of the pack. ChatGPT's letter packed the highest density of keywords pulled directly from the job description and scored best on the AI detector. However, that same density made parts of the letter read like a converted bullet list rather than natural prose, a pattern that cost it in tone and "AI tells" scoring.
Gemini
Here is the full Gemini cover letter ✍🏻
Chicago, IL | (555) 123-4567 | emailemail@email.com
Dear Hiring Manager,
With over five years of experience in digital marketing, demand generation, and marketing automation, I am writing to express my strong interest in the Mid-Level Marketing Manager position. My background in building multi-channel campaigns, optimizing conversion rates, and scaling organic pipeline directly aligns with your team’s strategic goals.
In my current role as Growth Marketing Specialist at Apex Cloud Solutions, I architect B2B demand generation strategies that have increased qualified pipeline by 38% year-over-year. I specialize in leveraging HubSpot and Salesforce to build automated lead nurturing workflows and lead scoring systems, which successfully improved our MQL-to-SQL conversion rate by 22%. Furthermore, I manage a $40k monthly paid acquisition budget across Google Ads and LinkedIn Ads, using continuous A/B testing to lower customer acquisition costs (CAC) by 18%.
My expertise spans both performance media and inbound content strategy. Utilizing Semrush and Google Analytics 4, I lead end-to-end SEO initiatives that drove a 65% increase in top-of-funnel organic traffic while building executive dashboards to monitor marketing ROI. Previously, at Vanguard Retail Tech, I executed targeted email marketing, social media campaigns, and landing page CRO experiments that achieved a 14% lift in site conversions. I also collaborated regularly with Product and Sales teams to align on go-to-market strategies and deliver impactful sales enablement collateral.
In addition to hands-on proficiency across HubSpot, Salesforce, Marketo, and GA4, I hold formal certifications in HubSpot Marketing Software, Google Analytics (GA4), and Google Ads Search. I bring a data-driven approach to funnel optimization and cross-functional leadership, with a consistent track record of connecting marketing operations to revenue growth.
I look forward to discussing how my experience in multi-channel campaign strategy and marketing automation can drive measurable acquisition results for your organization.
Sincerely,
Jane Doe
Gemini placed last overall. It was the only draft with a full contact header and kept a clean, readable layout throughout. However, it leaned heavily on generic marketing buzzwords, such as "leveraging," "impactful” and "end-to-end.” This is often enough that it read as noticeably more generated than the other two drafts, which pulled its tone and "AI tells" scores down.
Results Summary: Claude vs. ChatGPT vs. Gemini
| Model | Where It Won | Where It Lost |
|---|---|---|
| Claude | Natural, human-like voice and structure. Connects the candidate's skills directly to the job description's requirements without sounding like a robotic list. | Highest AI detector score: its formal syntax triggered a 100% flag rate on AI detection tools. |
| Gemini | Clean visual structure. Includes complete applicant header details and maintains a balanced, readable layout. | Frequent AI jargon. Overuses formulaic buzzwords like "leveraging," "impactful," and "end-to-end" that sound overly generated. |
| ChatGPT | Lowest AI detection rate (95.6%) and top keyword density. Maximizes tool and metric coverage while scoring slightly better on detection. | Predictable AI cadence. Reads like a laundry list of bullet points converted into stiff, mechanical paragraphs. |
Overall Ranking Breakdown
Claude finished first by winning four of the six criteria, producing the letter that read most like something a person would actually write and send.
ChatGPT took second on the strength of its keyword precision and its narrow win on the AI detector, even though its prose leaned mechanical.
Gemini came in third, as its polish and layout weren't enough to offset its buzzword-heavy phrasing, and it didn't top a single category outright.
One pattern stands out across all three: each letter scored above 95% on the AI detector. Regardless of which model "won," none of these drafts would pass as human-written without editing. This is a reminder that AI-assisted cover letters need a real human pass, not just a word count and keyword check, before they go out with an application.
How to Use AI for a Cover Letter Without Sounding AI-Written
While no model did a horrible job writing a cover letter, even the winning cover letter from Claude could benefit from some edits. Here are a few tips for using an AI job application tool to write the best cover letter possible:
- Start with your actual CV and the real job description, not a summary of either. This gives the bot as much information as possible to work from.
- Explicitly tell the model not to invent anything, including phrasing that implies a skill you don't have.
- Generate two or three drafts across different tools rather than accepting the first result.
- Rewrite any generic opening, unsupported claim or vague enthusiasm the model adds on its own. This includes editing out anything that could be seen as an AI “red flag,” such as the em dash or certain buzzwords. You may even want to ask the bot to explicitly avoid using these phrases or grammatical tools in your prompt.
- Conduct an AI cover letter review for factual accuracy and potential errors before you submit your application.
Once you have the first draft of your cover letter, consider following up with a review prompt, such as:
Review this cover letter as a hiring manager. Flag generic phrases, unsupported claims, awkward wording, and places where it does not clearly connect my experience to the role. Then suggest specific edits without inventing new experience.
How to Create Your Cover Letter Using Lorka
Lorka can help you craft a cover letter in multiple chatbots without ever opening a new tab. To get started:
1️⃣. Log in or create a Lorka account. You can also try Lorka for free here. Then, get your resume and the job description ready in new tabs.
2️⃣. Load the first bot of your choice, then feed it the following prompt:
Write a tailored cover letter for the role below using only the experience and achievements in the candidate's CV.
Requirements:
Keep the letter between 280 and 350 words.
Connect the candidate’s most relevant experience to the job description.
Use a professional but natural tone.
Do not invent experience, metrics, qualifications, employers, or skills.
Avoid clichés and generic enthusiasm.
End with a concise, confident closing.
Candidate CV: [paste here]
Job description: [paste here]
3️⃣. Paste this same prompt into 2-3 other chatbots of your choice. This will let you select the most usable cover letter, as we did in this experiment. You can change the bot in the lower-right corner of the chat.
4️⃣. Once the cover letters are complete, use the prompt from the tips section above to do an initial pass at the cover letter. As a reminder, that prompt is:
Review this cover letter as a hiring manager. Flag generic phrases, unsupported claims, awkward wording, and places where it does not clearly connect my experience to the role. Then suggest specific edits without inventing new experience.
5️⃣. Review your cover letter options and select the best one. Review it for any potential factual errors before submitting for an application. You can also use the tips from the section above to help polish off your AI-generated cover letter.
Why Use Lorka AI for Your Cover Letter Prompt?
Lorka allows users to create cover letters with multiple AI chatbots. So, you can find the best cover letter without ever having to load a new tab. You can try Lorka out for free using the GPT 4o mini. Alternatively, you can log into or create a Lorka account to use the cover letter prompt across a variety of popular AI models.
Verdict
There's no universal best AI for cover letters; the strongest draft depends on the applicant, the role and how carefully you edit the output afterward. In this test, Claude's letter needed the least cleanup overall, but how close second and third were is proof that "sounds the most natural" and "reads as least AI-written" aren't always the same model.
All three needed a human pass before submission. If you want to compare several AI cover-letter drafts side by side without switching between separate tools, Lorka lets you test the same application context across multiple models in one workspace.
Compare AI Models for Your Cover Letter
Generate your cover letter with Claude, ChatGPT, Gemini, and more—all in one place with Lorka AI.
Try Lorka AIFAQs
There isn’t one best option, though Claude won in this test largely for producing the most accurate and natural draft. The best bot for you will depend on a variety of factors, such as CV, role and prompt.
