Gemini vs Claude: Who Writes Your Emails More Naturally?

Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Artificial Intelligence at the Service of Email Writing
In the world of artificial intelligence, two tools stand out for email writing assistance: Gemini and Claude. While each has its own strengths, only one truly manages to capture the user's personal tone.
Gemini, developed by Google, is integrated into services like Gmail through the Help Me Write tool. This tool aims to enhance the efficiency of email drafting by considering the given context. However, after thorough testing, it turns out that Claude, the tool from Anthropic, excels in adapting to the user's style and personal requirements.
User Experience with Help Me Write
Three years ago, I started using Help Me Write, one of Google's first initiatives to integrate AI into its services. This writing assistant, present in Google Docs and Gmail, allows users to generate complete emails from simple prompts. While I appreciate this tool, I noticed that the results obtained through Google's standalone application, supported by the Gemini LLM model, were even more effective.
By providing Gemini with my email threads, I was able to compare its performance with that of Claude. After a week of testing, I was pleasantly surprised by Claude's ability to produce messages that seemed almost written by me.
Testing Methodology for LLM Models
To evaluate these competing models, I used Gemini 3.5 Flash, accessible via a Pro account, but also available for free with certain limitations. I activated the Personal Intelligence option from Google Labs to enrich the responses with stylistic cues drawn from my previous discussions.
On the Claude side, I used the free Claude Sonnet 4.6 model, set to a high effort level to balance speed and accuracy. I also utilized Claude's voice recognition, which proved superior, especially for accents.
For each email, I defined a clear objective and used two standard instructions to personalize the results and avoid wasting resources. First, I asked the AI not to respond immediately or ask follow-up questions. Then, I requested the model to mimic my tone in the email thread.
Notable Differences Between Gemini and Claude
From the outset, Gemini asked me two questions following my request. The first question was about the expected completion date, a relevant detail. However, the second question, regarding a response from the recipient, seemed redundant and lacked clear justification.
Claude, on the other hand, asked three questions, one of which was about the travel date, more pertinent for the recipient. Claude's second question demonstrated a better understanding of my instructions by directly asking for the desired firmness level in the email's tone. Finally, Claude extracted information from my email signature, although this detail was not directly related to the draft requested.
Contrasting Email Drafts
Using an entire email thread with an LLM requires considerable effort. Currently, typing my responses directly is faster than using AI, especially without automated integrations.
Gemini's response was complete but too lengthy, with a formal and friendly tone, inappropriate for a second request for information. Gemini's responses tend to be verbose, but this can be adjusted by disabling Personal Intelligence.
Example Response from Gemini
"I hope you are having a good week. I wanted to follow up on this since I haven't heard back from you regarding the tracking number. As mentioned earlier, a tracking ID or the name of the carrier would be incredibly helpful so that I can track the customs updates and ensure the package reaches me smoothly. Additionally, I wanted to inform you of my schedule. I will be traveling abroad in mid-July, so I am aiming for July 15 as the deadline to complete this product review. Given that we are already close to July, we are working with a tight window once the sample arrives. Could you please check back with Huanuo to get that tracking information? Once I know exactly when it arrives, I can officially schedule it in my content calendar and ensure we have everything wrapped up before my trip. Looking forward to hearing from you! Best, Chandraveer"
Example Response from Claude
"I hope you are doing well! I’m following up on the tracking number - I still haven't received anything from a shipper, so it would be great to have that detail. I also wanted to point out something to keep in mind: I will be traveling starting July 15, so I’d like us to have a well-established review timeline before then. Ideally, I want to have the unit in hand with enough time to spend some meaningful time with it and write something valuable. Could we agree on a delivery deadline and a review window that takes this into account? I don’t want us to be scrambling at the last minute. Looking forward to hearing from you - and I hope to hear from a carrier soon too! Best, Chandraveer"
Conclusion: Claude, a More Personal Choice
Although Gemini benefits from direct integrations with the Google ecosystem, it has not fully leveraged the context of emails. For Gmail users, Gemini may suffice, but its results are often verbose and imprecise. In contrast, Claude from Anthropic has demonstrated a superior ability to personalize responses through relevant questions and a better understanding of the desired tone. For those seeking greater customization, Claude seems to be the obvious choice.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.