As a contributing writer, my inbox often overflows, making efficient email management crucial. For years, I’ve relied on Google’s “Help Me Write” within Gmail and Docs for a quick assist with drafting replies.
While helpful, I’ve noticed the underlying Gemini LLM offers even more robust capabilities through its standalone app. This led me to wonder: could rival Anthropic’s Claude truly compete, or even surpass, Gemini in crafting emails that sound genuinely like me?
Setting the Stage: My AI Email Assistants
To put these AI powerhouses to the test, I dedicated a week to using both Gemini and Claude for my email replies. While both LLMs offer various models, I opted for specific configurations to ensure a fair comparison and optimal performance.
My setup for testing involved:
- Gemini 3.5 Flash: Accessed via a Pro account for speed, though it’s largely available for free with generous usage limits. Crucially, I enabled the “Personal Intelligence” toggle under Google Labs, allowing it to learn from my past Gemini chats for a more personalized touch.
- Claude Sonnet 4.6: I utilized the free version of this model, setting the “Effort” level to High. This choice aimed to balance speed with factual accuracy, mirroring Gemini Flash’s performance.
I typically dictate my prompts for efficiency, and it’s worth noting that Claude’s speech-to-text engine consistently outperformed Gemini’s, especially when dealing with accents.
My prompting strategy was consistent for both AIs to ensure comparable results and prevent unnecessary “token wastage.” First, I instructed them to withhold immediate responses and refrain from asking follow-up questions for details I hadn’t explicitly provided.
Second, and most importantly, I commanded the LLM to analyze and imitate my specific tone from the existing email thread. These two directives are essential for eliciting genuinely useful and personalized drafts from either AI.
The Initial Interrogation: AI’s Questions
To illustrate the process, I presented both AIs with a common scenario: drafting a follow-up email about shipment delays and a missing tracking number for a product I was scheduled to review. My prompt requested a new tracking number, explained my upcoming mid-July travel, and emphasized the need for a clear timeline to complete the review beforehand.
Here’s what I asked:
“Draft a response for the following email thread and request a shipment tracking number once again, explaining that since I will be traveling in mid-July, we need to put a timeline on this product review to ensure it is done before I travel. Do not respond immediately. Ask me any questions to draft a proper response. Imitate my semiformal tone seen in the thread.” (Followed by the email chain pasted as context.)
Gemini promptly responded with two questions. One asked for the targeted completion date for the review, which was a genuinely helpful detail I hadn’t specified. However, its second question, inquiring if the recipient had replied since my last email, felt redundant and lacked clear justification, especially since the existing thread already contained my prior tracking number request.
Claude, on the other hand, asked three questions, demonstrating a deeper “understanding” of my underlying needs. Instead of a review completion date, it smartly asked for my specific travel date, which is a more concrete and relatable detail for the recipient. Its second question directly probed how “firm” I wanted to be, explicitly stating this would influence the email’s tone—a level of transparency Gemini lacked.
Claude’s third question, pulling my website affiliations from my email signature, asked where the review would be published. While an interesting detail, it was ultimately irrelevant to drafting the immediate email and felt like a minor misstep, referenced from a prior, unrelated conversation.
Despite these slight variances, responding to these preliminary questions from both AIs was a significant effort. In the absence of direct integrations or automated scripts, simply typing out a response myself often proved faster than waiting for AI prompts, especially for non-Gmail accounts.
Drafting the Follow-Up: Who Got It Right?
With the preparatory questions answered, it was time to compare the actual email drafts generated by each AI.
Gemini’s response was three paragraphs long, covering all the bases. It clearly reiterated the tracking number request and highlighted my tentative submission date for the review. The tone was formal but leaned towards being friendlier than I’d prefer for a second urgent follow-up on the same unresolved issue.
I found Gemini’s draft to be somewhat verbose, which is a common trait I’ve observed in its outputs; its length diluted the urgency of my request. While you can sometimes mitigate this with “Help Me Write” or by disabling “Personal Intelligence,” for this specific scenario, the draft didn’t quite hit the mark.
Claude, leveraging its insightful pre-draft questions, delivered a more concise and impactful email, structured into just two paragraphs. Its language and tone were remarkably closer to my natural writing style, even cleverly incorporating a pun like “enough runway” when referring to my imminent travel schedule. This felt far more personal and urgent.
Beyond the superior draft, Claude also suggested a bundled subject line and, surprisingly, offered a “Send via Gmail” button. This level of integration, typically expected from Google’s own ecosystem, was an unexpected but welcome feature from Anthropic’s tool, especially since I logged in with my Gmail ID.
Claude’s final draft effectively conveyed the message:
“Hope you’re doing well! Just circling back on the tracking number – I still haven’t received anything from a shipper on my end, so it would be great to have that detail to work with. I also wanted to flag something worth keeping in mind: I’ll be traveling from the 15th of July, so I’d like to have a review timeline locked in well before that. Ideally, I’d want the unit in hand with enough runway to spend meaningful time with it and write something worthwhile. Could we agree on a delivery deadline and review window that accounts for this? I don’t want us to be scrambling close to the wire. Looking forward to hearing from you – and hopefully from a courier soon! Best, Chandraveer”
The Verdict: One Clear Winner for My Inbox
While Gemini undoubtedly benefits from its deep integrations within the Google ecosystem, it surprisingly fell short in this specific email drafting challenge. When using the standalone Gemini app, the results were often verbose and vague, not quite justifying the extra steps over simply typing an email myself.
In my rigorous testing, Claude consistently asked better, more relevant personalization questions, leading to superior and more human-like responses. Its ability to produce a concise, tonally perfect draft, complete with unexpected integration features, truly set it apart.
If you’re already relying on plugins and aiming for truly personalized, efficient email drafting, my experience points to Claude as the clear winner for crafting replies that genuinely sound like you.
Source: ZDNet – AI
