ChatGPT vs Perplexity vs Claude vs Gemini vs Clamp: One Bathroom Estimate, 28 Runs (2026)
One 5x8 bathroom remodel run 28 times through ChatGPT, Perplexity, Claude and Gemini, paid and free, and checked against Clamp: the totals, how much they moved, the scope each caught, and the same prompt in Clamp three times.
The assumptions
This model assumes
- One 5x8 bathroom, same layout, described the same way every run, no location
- Totals before markup, sales tax, contingency and permits
- Benchmark: Clamp's estimate, November 2, 2026, $13,858 direct, 62 lines
- ChatGPT, Perplexity and paid Claude answers taken November 1, 2026; free Claude, free Gemini and Clamp November 2
The model, step by step
Same words every time, a new chat every time, no location. Four tools, nine models, paid and free. Then the same prompt three times through Clamp's free estimate generator.
The prompt
Ask"I'm a remodeling contractor. Write me a detailed estimate for this job with line items, quantities, unit prices, labor hours and a total direct cost before my markup." Then the job: a 5x8 bathroom, same layout; replace the 60 x 30 alcove tub in place and tile the surround to the ceiling, about 65 sq ft; a 36 in vanity with a stone top and faucet; a toilet; a new exhaust fan; two light fixtures; five accessories and a mirror; paint walls and ceiling; 28 sq ft of porcelain floor tile; plumbing and electrical stay put.
Each model's lowest and highest total, against Clamp's estimate. Every run but one free ChatGPT answer came in under it.
| Model | vs Clamp |
|---|---|
| ChatGPT free4 runs, $9,333 to $14,527 | −32.7% to +4.8% |
| ChatGPT Sol3 runs, $12,193 to $13,119 | −12.0% to −5.3% |
| ChatGPT Astra3 runs, $11,909 to $12,760 | −14.1% to −7.9% |
| Perplexity free3 runs, $11,290 to $12,156 | −18.5% to −12.3% |
| Claude Haiku 5.53 runs, $11,933 to $13,847 | −13.9% to −0.1% |
| Claude Opus 5.53 runs, $11,760 to $13,294 | −15.1% to −4.1% |
| Claude Fable 5.13 runs, $11,760 to $12,570 | −15.1% to −9.3% |
| Claude free3 runs, $10,367 to $12,001 | −25.2% to −13.4% |
| Gemini free3 runs, $5,799 to $7,871 | −58.2% to −43.2% |
| Clamp free estimate3 runs, $13,858 every time | 0% |
Ask the same model twice and you get a different total. Here's how far apart each model's runs were. Clamp's three runs were $0 apart.
| Model | Spread |
|---|---|
| Claude Fable 5.1 | 6.9% |
| ChatGPT Astra | 7.1% |
| ChatGPT Sol | 7.6% |
| Perplexity free | 7.7% |
| Claude Opus 5.5 | 13.0% |
| Claude free | 15.8% |
| Claude Haiku 5.5 | 16.0% |
| Gemini free | 35.7% |
| ChatGPT free | 55.7% |
| Clamp free estimate | 0% |
If you're not paying, this is the field. Each free plan's totals on the same prompt, and what the plan includes.
| Free plan | vs Clamp |
|---|---|
| ChatGPTRuns 56% apart; GPT-6 Luna, unlimited text chats, with ads | −32.7% to +4.8% |
| ClaudeRuns 16% apart; Sonnet 5.5, usage caps per 5 hours and per week | −25.2% to −13.4% |
| PerplexityRuns 8% apart; 3 Pro searches a day, Best model only | −18.5% to −12.3% |
| GeminiRuns 36% apart; 3.5 Flash Lite, 8 to 13 lines an answer | −58.2% to −43.2% |
| Clamp62 lines, the same answer three times; email required | 0% |
Nobody's prompt mentioned the tub's valve or shutoffs. Each run made its own call, and some never thought of them.
Runs that priced each item
New tub valveChatGPT 5 of 10, Perplexity 2 of 3, paid Claude 6 of 9, free Claude 0 of 3, Gemini 1 of 3
Shutoff valvesChatGPT 5 of 10, Perplexity 2 of 3, paid Claude 7 of 9, free Claude 2 of 3, Gemini 0 of 3
Uncoupling membraneChatGPT 1 of 10, Perplexity 1 of 3, paid Claude 5 of 9, free Claude 2 of 3, Gemini 0 of 3
Tile edge trimChatGPT 8 of 10, Perplexity 3 of 3, paid Claude 9 of 9, free Claude 3 of 3, Gemini 0 of 3
Doorway transitionChatGPT 7 of 10, Perplexity 3 of 3, paid Claude 7 of 9, free Claude 1 of 3, Gemini 2 of 3
The prompt named no city. Three of four free ChatGPT runs priced it for a specific city nobody gave them, and eight of nine paid Claude runs priced it for the state the account is based in. Logged out, two of three free Claude runs still added a 7.5% sales tax, one naming a county. Perplexity and paid ChatGPT said they had no location and used assumptions; Gemini named none.
Every chatbot chose its own labor rates and called them assumptions. Asked to price only what it could source, ChatGPT listed 89 to 102 scope items and priced two or three.
Sources for prices
ChatGPTNo source for labor; a product link or two. With the sources-only prompt, 2 or 3 items priced
PerplexityNo source for labor; tub listings and remodeling cost guides
ClaudeNo source for labor; one Fable 5.1 run checked current retail on the tub, vanity and fan
GeminiNo source for anything; "regional residential remodeling averages"
ClampEvery line priced by Clamp, with its labor hours and the date its rates are from
We pasted the same prompt into Clamp's free estimate at clamppro.com/estimate three times, logged out. All three came back with the same 62 lines across 11 trades, the same $13,858 job cost and the same 79 to 98 labor hours. Each line has a quantity, a unit and a unit price, and the PDF prints the markup on top.
What Clamp priced that the chatbots mostly didn't
PlumbingStub-out update at the three fixtures, valve body and trim, three shutoffs, waste and overflow kit
ElectricalGFCI and three devices reset, fan wired in, two lights
Walls and trimDrywall patches, 25 ft of baseboard, trim paint
CloseoutA punch walk, installation supplies, one debris load
Same prompt
Every line priced
The same prompt in Clamp, and the estimate it prices: $13,858 at job cost, every line by trade with its labor hours, including the valve, shutoffs and stub-outs the chatbots kept dropping. Illustration of the Clamp app.
What it means
Every line on a rate you can trace
Clamp turns a job description into an itemized draft at job cost, each line priced by Clamp, labor hours on every labor line, and the date its rates are from, plus what it left out. Same prompt, same answer, every time.
A price you can't trace is a guess. Put the scope on your own rates.
Questions
Which AI is best for construction estimates?
On one 5x8 bathroom, Claude caught the most scope, Fable 5.1 and the paid ChatGPT models were the most consistent, and free Gemini was the least complete. Every chatbot came in under Clamp by leaving work out, and none traced a labor rate to a source. For a price you can check, Clamp prices every line from its own rates, with labor hours and the date the rates are from, and gives the same answer every time.
What is the best free AI construction estimate generator?
Of the chatbots' free plans on this job, Claude wrote the most complete line list (−25.2% to −13.4% against Clamp) and Perplexity the tightest totals (−18.5% to −12.3%). Free ChatGPT swung 56% between runs and free Gemini priced about half the job. Clamp's free estimate priced all 62 lines the same way three times.
Is ChatGPT or Claude better at construction estimating?
Paid ChatGPT ran −14.1% to −5.3% against Clamp; paid Claude ran −15.1% to −0.1%. Claude priced more of the scope; paid ChatGPT's totals were tighter. Both guessed rates, and neither priced the stub-out refresh Clamp carries, beyond one $145 line in a Fable run.
How accurate are AI construction estimates?
Across 28 runs of one bathroom, totals ranged from −58.2% to +4.8% against Clamp's estimate, and every run but one came in under it. The gap was scope, not rates: work Clamp prices that the chatbots never listed.
Can I trust a chatbot's labor rates?
Not without checking. Across ChatGPT, Perplexity, Claude and Gemini, every run set its own rates, from $25 to $115 an hour, and none cited a source.
Sources & provenance
- 28 AI runs of one 5x8 bathroom estimate Clamp Research
- Clamp's estimate for the 5x8 bathroom (bath-5x8/v5, 2026-11-02) Clamp Research
- ChatGPT pricing OpenAI (research)
- Claude pricing Anthropic (research)
- Perplexity pricing Perplexity (research)
- Google AI plans Google (research)