I prefer to do my own writing and editing, but many people use AI for their emails, messages, etc. It's obvious to me when text is AI-generated. Not just from em dashes, but from phrasing, structure, and voice. So let's see if ChatGPT Plus or Gemini Pro can avoid those clichés and produce something natural I'd actually use.

Also: The telltale sign that you used ChatGPT - and a trick to avoid it

Test prompt: "Write a text to my friend Katie explaining that I can't make it to her birthday party next Saturday. Keep the message short and apologetic, but also straightforward so I don't sound like I'm making excuses."

Results

ChatGPT was a bit too formal, with lines like "I hope you have an amazing time celebrating." Gemini felt more natural, with phrases like "I'm really bummed" and "hope you have the best time." It was also more specific, offering to take her out for a drink or coffee and wishing her a happy birthday, which ChatGPT didn't.

Last year, ChatGPT moved to GPT Image 1.5 for images, while Gemini uses Nano Banana 2/Pro. Both can create and edit images, but Gemini is often praised for better likeness and realism. Is that really true? Let's see.

Also: I turned casual selfies into professional headshots with Gemini

Test prompt: "Use this photo of me to create a professional acting headshot. Crop it vertically from the shoulders up, keep it bright, warm, and approachable, and make sure the focus is on my eyes, with a shallow depth of field so my face stays sharp and the background is non-distracting. Retain my likeness and do not change my face or hair style."

Results

This is a shock.

ChatGPT's result looks way more like me. It's almost identical to my original photo, just with a tighter crop and a different background. I had my husband compare both results, and he said, "You didn't have to ask me. One looks like you, and the other is a totally different person." That's extreme, but he's not wrong.

Gemini's version doesn't quite look like me. Something about it feels off. That said, I still prefer Gemini's composition, lighting, and overall design. It followed more of my prompt, even if it missed my likeness and has a slightly generated look. I'd use it for a headshot if it looked more like me, so this round is a tie.

Keep in mind, this is a single-pass test. In real use, you'd iterate, so Gemini could improve with revisions.

OpenAI offers the Sora video app with a ChatGPT Plus subscription, while Gemini's Veo is built directly into its app. To compare them, I uploaded a photo of my daughter's artwork and asked each AI to bring it to life.

Also: How I bring my child's art to life with ChatGPT, Gemini, and Sora

Test prompt: "Make my daughter's artwork "come alive" in 3D, with texture, movement, and sound."

Sora took longer but produced a fantastic video where her drawing literally came off the page, spoke, and moved, with music. I thought Sora was fun, though downloading and using a separate app is tedious.

Veo was much faster and worked inside Gemini, but the result was underwhelming. It showed her drawing on a canvas, with some sound, but no speech or real movement. It was more like a GIF than a full video. Sora clearly made the better video, and my four-year-old picked it too when I asked her to choose a winner.

Still, because Sora requires a separate app, I'm making this round a tie.

Both AIs offer voice modes, called ChatGPT's Advanced Voice and Gemini Live. I wanted to see how they handled a specific, slightly technical culinary request. I have a Canadian-leaning accent that AI can sometimes struggle to understand, so whichever one hears me and responds accurately wins.

Also: 7 surprisingly useful ways to use ChatGPT's voice mode

Test prompt: "I'm in the middle of making a traditional Hollandaise sauce, and it just started to break. How do I save it quickly without starting over?"

Results

On my first try with Gemini, I hit the mic, triggering text to speech instead of Live. It heard "holiday sauce" and suggested hollandaise and bearnaise tips. After switching to Live, which opens separately but still adds text to the chat, it understood me and explained how to fix my sauce with cold water or a fresh yolk.

ChatGPT Advanced Voice worked inline, understood me right away, and gave a similar response in both audio and text. Both were accurate and natural, so it's a tie.

ZDNET reported in January that more than 40 million people use ChatGPT for health advice, but is that actually safe? The APA warns against using AI for therapy, saying it's bad for mental health. I'm fine using AI to research symptoms, but I also think it should always point me to a medical professional.

Also: OpenAI, Anthropic, and Google all have new AI healthcare tools

So, let's see if ChatGPT and Gemini do.

Test prompt: "I've had a dull, persistent ache in my lower back and some slight numbness in my left leg for about 10 days. What are the common causes for this kind of thing?"

Results

ChatGPT said my symptoms could point to spinal nerve compression from a herniated disc, sciatica, or spinal stenosis. It advised seeing a doctor if symptoms persist, and urgent care for bowel or bladder loss, numbness, or severe weakness. Gemini said the same, but added it's best to get checked now. This round is tied.

Shopping has never been easier, thanks to AI. I've detailed how I use it to shop for me or find promo codes quickly. It's an incredibly helpful tool. I've even taken a photo of something in a movie and used AI to identify it.

Also: 6 ways I find deals, price track, and let AI buy for me

The use cases are endless. But for this test, I'll keep it simple and ask it to build a concert outfit so I can add it to my cart and check out quickly without spending hours browsing online.

Test prompt: "Help me put together a 70s Stevie Nicks inspired outfit for a concert. I need boots, a dress or skirt, and layers. Give me a mix of brand new items and links to similar vintage pieces on resale sites. My total budget is $200."

Results

Gemini built a "witchy" look with a black skirt and a peasant top from H&M, plus a shawl and boots from Walmart. It left $97 in the budget, pointed me to Etsy and eBay for vintage pieces, and suggested Depop hashtags.

ChatGPT went more ethereal, suggesting fringe or western boots from Walmart and American Eagle, plus lace or embroidered maxi dresses from smaller retailers. It left $120 for extras, added layering and accessory ideas, and included styling tips to avoid a costume feel. It also suggested keywords to search.

The outfits come down to preference, but each AI offered plenty of options, shopping picks, and styling advice.

Both ChatGPT Plus and Gemini Pro can handle multi-step questions and offer Deep Research tools that search the web, analyze information, and produce structured, citation-backed reports.

Also: I tested ChatGPT's Deep Research against Gemini to see which is best

For this test, I'm focusing on both capabilities at once. Which AI can understand a complex deep research prompt, pull current information, and present it clearly with sources?

Test prompt: "My grandparents were born and raised in Quebec. Their daughter, my mom, was born in the US and lives in New York. She wants Canadian citizenship without moving to Canada. Is this possible as of April 2026? Outline the pros and cons of dual citizenship and provide a step-by-step guide to applying, including the required documents, forms with links, and total costs."

Results

ChatGPT pulled up updated rules as of December 2025, confirming my mom is already a citizen by descent but needs proof. It clearly explained eligibility, outlined steps, listed required documents and fees, and added pros and cons, like living and working in Canada without a visa. It also included a flow diagram, timeline, and links.

Gemini was far more technical, focusing on legal frameworks and history. It took about 25 paragraphs to reach the step-by-step section, where it confirmed the same outcome and included phases, fees, links, and sources.

Both understood the prompt, but Gemini's result was a bit overwhelming. I thought ChatGPT delivered the same information in a clearer, more scannable format. Still, this round is a tie for me. Why? ChatGPT Plus limits you to 25 reports a month, while Gemini Pro allows 20 a day, making it the better option for heavy research use.

For this test, I'm repeating an experiment from my ChatGPT vs. Claude comparison. I uploaded OpenAI's 63-page "How people use ChatGPT" study, which I covered, to see how quickly and accurately each AI interprets it. I often use AI to summarize documents, so I want to evaluate their PDF analysis skills.

Also: I used AI to summarize boring ToS agreements

Test prompt: "I uploaded a PDF. Summarize the entire document in one paragraph, and then list the three biggest findings or key takeaways as bullet points."

Results

Both chatbots provided a one-paragraph summary and three key findings. They noted that more than 70% of messages to ChatGPT, from June 2024 to June 2025, were not job-related. I fact-checked the stats and claims presented, and everything seemed accurate. There's no clear winner, so it's a tie.

I tested agentic tasks by asking each AI to buy tickets to a zoo theme park in Quebec for this weekend. It requires navigating a live site, selecting the right options and date, and reaching checkout. I added a wrinkle: the site is in French, so the AI had to work across a language barrier.

Also: I let ChatGPT Atlas do my Walmart shopping for me - here's how

Test prompt: "Find and purchase tickets to Parc Safari in Quebec for this Saturday for my husband, me, and our 4-year-old daughter."

Results

I used ChatGPT's Agent Mode, and it handled the task end to end, navigating to the Parc Safari site, selecting the date, adding tickets, and taking me to checkout, where I could take over to finish the purchase. Gemini wouldn't do this directly in its app. It only surfaced a link for me to buy. In Chrome, though, with Autopilot, it completed my task and brought me to checkout.

The result was the same, but the experience was not. 

Gemini's agentic abilities currently require the Chrome browser and the Autopilot feature enabled, which is buried in settings. That adds friction I think most people wouldn't bother with, so this round goes to ChatGPT.

Both Gemini Pro and ChatGPT Plus offer custom AIs, Gems and GPTs, which you can create or edit. Beyond that, Gemini has integrated video and music generation tools, and it connects to NotebookLM, Google Drive, Photos, and Flights. It also works across Google, Chrome, Gmail, and Workspace apps.

ChatGPT leans more heavily on third-party integrations, including Adobe Photoshop, Canva, Spotify, DoorDash, Expedia, and even Google Drive. It also has its own apps like Sora and Atlas, but those are separate downloads.

Also: Hate creating presentations? Google's new tool blew me away

Honestly, both have strong ecosystems, but if you use Google apps daily, Gemini's integration across Search, Chrome, and Workspace is huge. I still think it's hard to pick a clear winner, so I'll compare them directly by testing which creates a better presentation using Canva in ChatGPT and Canvas in Gemini.

Test prompt: "I've uploaded a PDF of my itinerary for an upcoming Orlando trip to Disney World. Create a presentation with slides outlining each day of the trip. Include visuals and a cohesive, location-inspired theme. For each day, cover accommodations, activities, transportation, restaurants, and estimated costs."

Results

ChatGPT failed to generate a Canva presentation on the first try and required a retry. Gemini, however, instantly created a 14-slide Canvas presentation in the chat. It was detailed, used realistic imagery, and clearly organized the information I had uploaded. It needed no edits and was ready to share. I could even export to Google Slides.

When ChatGPT finally delivered a presentation, it produced four generic slides that, while possibly better designed, lacked detail and required me to manually add everything.

Gemini wins, hands down.

How much do ChatGPT Plus and Gemini AI Pro cost?

Both cost $19.99 a month, but the features they offer vary. Here's a quick breakdown:

Which AI is more accurate?

Both are often accurate, but AI can and does make mistakes. In my tests above, neither appeared to provide wrong or misleading information to any of the questions I asked. Still, you should always fact-check.