The AI Tool Shelf- We gave ChatGPT, Gemini and Claude the same five tasks on their free plans in October 2026.
- All three searched the web for news, but each picked a completely different set of three stories.
- ChatGPT opened the profit calculation with a wrong answer before correcting itself; Gemini and Claude were right from the start.
- For Standard Arabic our native speaker ranked Claude first, Gemini second and ChatGPT third; all three wrote natural Moroccan Darija.
ChatGPT, Gemini and Claude are the three most-used AI assistants, and all three have a free plan. Instead of comparing feature lists, we gave each one the same five tasks on its free plan: a YouTube description, this week's AI news, a business calculation, a paragraph in Modern Standard Arabic and a message in Moroccan Darija. Here is what each one did, with screenshots, and which ones we would use.
How we tested
- Free accounts on chatgpt.com, gemini.google.com and claude.ai, on October 24, 2026, with each tool's default model.
- The exact same prompt pasted into each tool, one task per conversation.
- We noted whether each answer followed the instructions, got facts and numbers right, and sounded natural.
The results at a glance
| Task | ChatGPT | Gemini | Claude |
|---|---|---|---|
| 1. YouTube description (under 120 words, 3 hashtags) | ✅ Followed both rules | ✅ Followed both rules | ✅ Followed both rules |
| 2. This week's AI news, with dates and sources | Searched the web; 3 stories with links, but one was from before this week | Searched the web; 3 stories with dates and sources | Searched the web; 3 stories with links, but answered in Arabic |
| 3. Profit calculation | ⚠️ Opened with the wrong answer, then corrected itself | ✅ Correct ($49.60, small rounding) | ✅ Correct ($49.66), plus margin advice |
| 4. Modern Standard Arabic | Good (3rd) | Very good (2nd) | Best (1st) |
| 5. Moroccan Darija | Natural | Natural | Natural |
Task 1: a YouTube description
All three passed easily: each stayed under 120 words, added exactly three hashtags and ended with a call to subscribe or comment. Gemini's was the most detailed, mentioning the "translation workflow" and a side-by-side comparison; ChatGPT's was the most classic YouTube style; Claude's was the shortest and added two questions to invite comments. Claude also added a short note in Arabic, because it remembered from earlier chats that we usually write in Arabic. Honestly, any of the three would do the job here.
Task 2: this week's AI news
This is where things got interesting. All three searched the web on the free plan and gave dated stories with sources, but they picked three completely different sets of stories:
- ChatGPT: OpenAI resuming advanced training after a security overhaul (Reuters), Anthropic donating its Agent Skills standard to an open foundation, and China's proposed AI export controls. The answer was the most polished, with images and "read the report" links, but the China story was dated October 16, before "this week".
- Gemini: an Anthropic and UK AI Safety Institute report on unsanctioned AI behaviour, a critical Node.js vulnerability found by OpenAI researchers, and Meta's Muse app passing 10 million downloads.
- Claude: the PHP2Shell vulnerability found with AI, OpenAI models coming to Google Cloud, and Moonshot's Kimi K3.1 Pro. It flagged one date as approximate, but answered in Arabic even though the question was in English, again because of its memory of our earlier chats.
The lesson: "the most important news" is a judgement call, and each assistant searches and ranks differently. For news, check the sources, and don't rely on one assistant alone.
Task 3: a business calculation
The correct answer: 40 sales, $4.24 net per sale, about $169.66 in revenue and a profit of about $49.66 a month.
- ChatGPT opened with a bold "You lose $24.66 per month", then worked through the numbers, reached $49.66 profit and added a "Correction" at the end. The final answer was right, but a reader who stopped at the first line would have been misled.
- Gemini got it right ($49.60, after rounding Google's fee to $0.75).
- Claude got it right ($49.66) and added useful context: you earn about $1.24 per sale, so the business loses money if each sale costs more than about $4.24 in ads.
Always read the whole answer, and check any number you plan to act on.
Task 4: Modern Standard Arabic
All three wrote correct, fluent Arabic. Our native-speaker ranking: Claude first, for the most natural and well-structured paragraph, which also reminded the owner to review sensitive replies himself; Gemini second, very good and ending with an offer to draft ready-made replies; ChatGPT third, correct but a little more generic.
Task 5: Moroccan Darija
Darija is a real test: it is a spoken dialect with no standard spelling, and many AI tools drift into Standard Arabic or Middle Eastern dialects. Here, the result surprised us: all three wrote natural Moroccan Darija with good, authentic words, and all three added light humour about the food and the dancing. Gemini and Claude offered two versions (a longer and a shorter one); ChatGPT gave one with quick buttons to shorten it or add the time and place.
For Arabic-speaking and Moroccan users, this is good news: all three free assistants can now handle everyday Darija well.
Free plan differences we noticed
- ChatGPT showed sponsored suggestions under some answers on the free plan, clearly labelled as ads.
- Claude used its memory of our past chats, which helped with tone but also made it answer in Arabic when we asked in English.
- Gemini often ended with a helpful follow-up question or offer.
- All three searched the web for free when the question needed current information.
Which one should you use?
- Claude was the most reliable in our test: correct maths on the first line, the best Standard Arabic and natural Darija. It is the one we would pick for writing and careful work.
- Gemini was a close second: accurate, fluent in Arabic and Darija, and a natural fit if you live in Google apps.
- ChatGPT gave the most polished, visual news answer, but its calculation started with a wrong conclusion and its Arabic ranked third for us.
After this test, the two we would keep using are Claude and Gemini. All three are free to try, so the best move is to run your own most common task in each one. For plans and prices, see our overviews of ChatGPT, Gemini and Claude, or the guide to the best free AI chatbots.