
Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- Gemini powers Help Me Write for efficiency in Gmail.
- You can use Gemini or Claude to draft context-aware emails.
- Testing showed Claude is better at matching tone and requirements.
Three years ago, I started using Help Me Write, one of Google’s first attempts at applied AI. The writing assistant is built into Google offerings like Docs and Gmail, and I found it could draft whole emails from scratch — all I had to do was sign off on them. As much as I like Help Me Write, I’ve noticed the Gemini LLM underpinnings powering it somehow deliver even better results through Google’s standalone AI app.
Since I was feeding Gemini my email threads for drafting responses, I decided to compare Claude’s performance. After a week’s worth of testing, I surprisingly found Anthropic’s tools write messages a lot as I would.
What I’m comparing, and how I’m testing
These rival LLMs offer multiple models, but I’m sticking to Gemini 3.5 Flash for speedy response times, accessed through a Pro account. However, it’s available for free, with liberal usage limits. Importantly, I’ve switched on Google Labs’ Personal Intelligence toggle under Attachments > More Tools > Personal Intelligence. It pulls actionable stylistic cues and information from my prior Gemini chats, adding a personal flair to the current responses.
As for Anthropic’s tool, I leveled the playing field with the free Claude Sonnet 4.6 model, and Effort
set to High. It strikes a balance between speed like Gemini Flash, and all-important factual accuracy. Additionally, I dictate my prompts to save time, and Claude’s speech-to-text engine is leagues ahead of Gemini, especially if you have an accent.
Also: I let Gemini Flows organize my Gmail, and it effectively filtered my inbox (with one sneaky catch)
For every email, I outline the draft’s objective, then follow with two standard instructions. They personalize results consistently and prevent token wastage. First, I ask the AI to suspend immediate responses and refrain from follow-up questions seeking details missing from the initial prompt. Second, I command the LLM to imitate my tone from the email thread. You’ll need these to expect a workable response.
To illustrate, I’ll use an email following up on shipment delays and a missing tracking number for a product I plan to review.
Here’s the prompt:
“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.
“<Email chain pasted/attached as PDFs>””
Differences show immediately
Right off the bat, Gemini asks me two questions following up on the request. The first seeks a legitimately useful detail I didn’t mention — the targeted completion date for the review.
Sadly, Gemini’s second question asks if the recipient has replied since my last email in the thread, but doesn’t state why it needs my answer. Questioning the supplied info seems redundant since I would’ve obviously included the latest response, and its accompanying justification remains ambiguous. The email thread includes a prior request for the tracking number, which the AI ignored.
“If he did reply to your request for the tracking number, knowing what he said will help me address his latest points directly. If he hasn’t replied and you are following up again, I will frame it as a gentle nudge. “
Also: Your Gmail is getting an AI makeover – here’s what to expect and when
Claude performs similarly, but asks three questions instead. While Google’s tool asked for my planned review completion date, Anthropic requested the specific travel date, which is arguably more relevant and relatable information for my recipient. I cannot reasonably commit to a completion date while still awaiting the product, but the travel date is a fixed constant.
The second question highlights Claude’s superior “understanding” of my prompts, directly asking me how firm I want to be. The justification explicitly states my response will influence the email’s tone, unlike Gemini’s vagueness. Third, Claude pulled the websites I write for from my email’s signature line, and asked me where I would write the review in question.
This detail is irrelevant to the requested email draft and was referenced in a prior conversation, leaving me mildly disappointed. To keep testing fair, I responded to all the questions from both AIs.
The email drafts are leagues apart
Since this approach of including an entire email thread with an LLM’s prompt takes considerable effort without integrations or scripted automations, I’d say my expectations were lofty. Currently, typing out responses directly is just faster than using AI or improvising something from Help Me Write in Gmail, and I don’t even get that convenience in non-Gmail accounts.
Gemini’s response covered all the bases, but ran three paragraphs long. They were structured well, with my tentative submission date in bold. The tone was formal, but friendlier than I’d prefer using in the second request for the same info.
I’ve noticed that Gemini’s responses are typically verbose, but you can dodge that with Help Me Write, or switch Personal Intelligence off when you don’t need to reference data from other chats. I found that the length of this email draft diluted the urgency too much.
Gemini’s response:
“Hope you’re having a good week.
I wanted to circle back on this since I haven’t heard from you regarding the tracking number. As mentioned before, a tracking ID or courier name would be incredibly helpful so I can keep an eye out for customs clearance updates and ensure the package reaches me smoothly.
Additionally, I wanted to give you a heads-up regarding my timeline. I will be traveling out of the country in mid-July, so I am targeting July 15th as a hard deadline to wrap up this product review.
Because we are already close to July, we are working with a tight window once the sample arrives. Could you please check in with Huanuo once more to secure that tracking info? Once I know exactly when it’s landing, I can officially lock this into my content calendar and guarantee we get everything done before I travel.
Looking forward to hearing from you!
Best, Chandraveer”
Also: Anthropic rolls out Claude Tag, your new agentic AI coworker in Slack
With Claude, the personalization and relatable cross-questioning laid the groundwork for a well-rounded draft. Like Gemini, it didn’t omit facts, but it produced just two paragraphs, closer to the length I would’ve written. Most importantly, the tone and word choice were splendid. While Google just spat out a draft, this AI highlighted the included pun “I’d want the unit in hand with enough runway to spend meaningful time” referring to my imminent travel.
A bundled subject line and a Send via Gmail button were the cherry on top. The latter appeared since I logged into Claude with a Gmail ID, but it’s something you’d expect of Gemini instead, given its direct ecosystem integration.
Claude’s response:
“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”
My new preferred tool
It’s no secret that Gemini benefits from direct integrations with Google’s far-reaching ecosystem, but it dropped the ball, especially given the prompt’s clear correlation to emails. If you’re using a Gmail account, the Gemini LLM gets the job done with Help Me Write, but the results are verbose and vague. Using Gemini the roundabout way isn’t worth the hassle.
In my tests, Claude routinely asked better personalization questions to improve responses. Using it is simplified through integrations enabled in the Skills section. So, if you’re plugin-dependent anyway, and picking between the two, I’d pick Claude.
