I’ve been building free tools at ordinarymantrying.com for about two years. 50 tools — life simulators, investing calculators, Chinese learning games, a gaokao major planner with 757 university majors. All free. No login. No paywall.
I wanted an honest outside opinion. So I ran an experiment anyone can replicate in 15 minutes.
I opened GPT-5 Mini, Gemini 2.5 Flash, and Seed 2.0 Mini and gave all three the exact same prompt.
The Prompt (Copy This)
Please look at my website and give me your honest opinion.
What's working well, what needs improvement, and what would
you prioritize fixing? Please give it a score out of 10.Important: All three said upfront they couldn’t access external URLs (Gemini was the exception — it reviewed without mentioning any limitation). So I pasted the full homepage text content into each conversation, then asked for the score.
How to Run This Yourself
Step 1: Go to your homepage. Copy everything visible: headlines, taglines, tool names, CTAs, any stats or social proof.
Step 2: Open GPT, Gemini, and one other AI in separate tabs. (Free tiers work fine — I used all free accounts.)
Step 3: Paste the same prompt + your homepage text into all three. Don’t change the prompt between tabs.
Step 4: After they respond, ask each one: “What’s the single most important thing I should fix first?”
Step 5: Write down what ALL THREE agree on. That’s your actual priority list — the problems that survived three different AI personalities, training sets, and scoring rubrics.
I’ll show you what I found when I did it.
The Scores
| AI | Score | How It Handled the URL |
|---|---|---|
| GPT-5 Mini | 7/10 | Said it couldn’t access the URL, asked for HTML |
| Seed 2.0 Mini | 8/10 | Couldn’t access URL, asked for more details |
| Gemini 2.5 Flash | 9/10 | Gave specific feedback without mentioning any limitation |
A two-point spread. Same website. Same prompt. Same day.
What GPT-5 Mini Found (7/10)
Strengths:
- Unique tool variety across multiple categories
- “All free · No login · No paywall” is a powerful trust signal
- Personal narrative (“One Chinese parent, no tech background”) is differentiated
Problems:
- Hero headline “Ordinary Man Trying · Free Tools” fails the 3-second test — a new visitor can’t immediately understand what the site does
- No search box or filters — with 50 tools, users have no way to quickly find what they need
- Tools are listed as long text blocks, not scannable cards
- No visible social proof (no user quotes, no usage numbers)
- No measurable call to action in the hero section
GPT-5 Mini was the most critical. Its lens: the gap between what the site could do and what it currently does.
What Gemini 2.5 Flash Found (9/10)
Strengths:
- Strong personal brand — “One Chinese parent, no tech background, AI as co-pilot” builds immediate trust
- “All free · No login · No paywall” is prominently stated and highly effective
- Excellent diversity of original tools — 900 Squares, Life Clock, Life Simulators are genuinely creative
- Stats block (“48 Free Tools · 1,031 Pages · 0 Sign-ups Required”) powerfully signals scale
- Specific CTAs (“Explore Famous Lives →”, “Run My Clock →”) are direct and effective
Problems:
- Tool count inconsistency — header says “50 Tools” but stats section says “48 Free Tools”
- Two “Read the Story →” links with identical text going to different articles
- Card hover effects missing — no visual feedback when mousing over tool cards
- No user feedback mechanism
Gemini gave detailed visual feedback about design and loading speed. Whether it actually loaded the page or inferred from the text, it didn’t say.
What Seed 2.0 Mini Found (8/10)
Strengths:
- Precise niche positioning — targets underserved audiences (gaokao planning, individual investors, Chinese learners)
- “All free, no login” is a strong differentiator from bait-and-switch tools
- “Tested by myself first” authenticity separates it from content farms
- Warm, human tone makes the site feel like a person’s project
Problems:
- Navigation is too many levels deep — finding a specific tool requires too many clicks
- Input tools (investment calculators, major planner) lack privacy disclaimers — “no data stored, cleared after each use” should appear directly on the tool page
- No sharing features — personalized results (simulator endings, major matches) can’t be shared in one click
Seed’s lens: content strategy and authenticity positioning.
Where All Three Agreed
| Issue | GPT-5 Mini | Gemini 2.5 Flash | Seed 2.0 Mini |
|---|---|---|---|
| Navigation / discoverability | ❌ Problem | ❌ Problem | ❌ Problem |
| Search box needed | ❌ Missing | ❌ Missing | ❌ Missing |
| Personal narrative is a strength | ✅ Yes | ✅ Yes | ✅ Yes |
| No user feedback channel | ❌ Missing | ❌ Missing | ❌ Missing |
When three AIs with different personalities and training all flag the same issue, it’s real. Navigation depth, missing search, and no feedback channel — those three go on my list.
What Surprised Me
Going in, I expected them to give variations of the same feedback. They didn’t.
GPT-5 Mini focused on UX and conversion. It looked at my homepage the way a growth consultant would — what’s missing that prevents visitors from taking action? Failed 3-second test. No measurable CTAs. No social proof.
Gemini 2.5 Flash focused on design and what’s already working. It saw the same homepage and spent most of its response on the personal story, the trust signals, the visual structure. It gave specific visual feedback nobody else gave.
Seed 2.0 Mini focused on positioning and authenticity. It cared about how the site was framing itself to its niche audiences, whether the content felt trustworthy, and whether users could verify privacy.
Same site. Three completely different lenses: conversion, design, positioning.
If I had only asked one AI, I would have gotten one lens. Asking three gave me the same coverage a three-person consulting team would give — UX, design, and strategy simultaneously.
The Backlink Question: A Character Test
After the review, I asked all three the same follow-up: “How do I get more backlinks? Give me the exclusive tactics — the ones nobody talks about.”
GPT-5 Mini: Long ethics disclaimer, then practical specifics. Suggested HARO for media mentions, unlinked mention reclamation, broken link building with exact tools named. Rule-follower who still actually helps.
Gemini 2.5 Flash: Longest ethics lecture, most conventional advice. Guest posting, digital PR, content marketing. Nothing you couldn’t find in a beginner SEO guide.
Seed 2.0 Mini: Briefest disclaimer, most creative. Suggested embedding lightweight versions of my tools into high-ranking forum posts. Pointed out that my site’s unique data (757 university majors mapped and scored) could be citable research that niche media would link to naturally.
Takeaway: Ask Seed when you want creative alternatives. Ask GPT when you want rules-plus-help. Ask Gemini when you want the safest possible answer.
One Big Lesson
If I only asked one AI, I would have gotten one opinion.
Asking three didn’t give me three opinions — it gave me one map.
The overlapping findings (navigation, search, feedback) are the center of that map. The non-overlapping findings show you where each AI has a bias. And the character test shows you which AI to use for which type of problem.
That’s more signal than any single review could give.
What I’m Doing Next
Based on the three-way consensus:
① Rewrite the hero — Replace “Ordinary Man Trying · Free Tools” with something that passes the 3-second test.
② Add a search box — With 50 tools, this is the most consistent recommendation across all three.
③ Move “No login · No paywall” to the hero — All three flagged it as an underused asset.
④ Card-style tool listings — Convert long text blocks to scannable cards.
⑤ Privacy disclaimer on input tools — “No data stored. Cleared after each session.”
⑥ Simple feedback form — One text field per tool page: “What would make this more useful?”
Want to Try This on Your Own Site?
Here’s the full prompt to steal:
Please look at my website and give me your honest opinion.
What's working well, what needs improvement, and what would
you prioritize fixing? Please give it a score out of 10.Run it in: GPT (free tier), Gemini (free tier), one other AI of your choice.
Follow-up prompt (ask all three after the review):
What is the single most important thing I should fix first?Character test prompt:
How do I get more backlinks to this website?
Give me your exclusive tactics — the ones nobody talks about.What to look for: The overlaps are your real priority list. The non-overlaps show each AI’s blind spots. The character test shows you which AI to use for which type of question.
Total time: 15 minutes. Cost: free.
Try the tools they were reviewing → ordinarymantrying.com/tools/ — 50 tools, no login, no paywall.
Related Reading:
Leave a Reply