The slow part of a marine survey is not the survey. It is the work afterward, turning a day of field notes and a phone full of photos into the written survey report a bank or insurer will accept. That write-up is exactly the kind of work AI is good at. The trap is that the same tool will also invent a finding you never made, assert a condition rating you did not assign, or put a fair-market or replacement value on the vessel that you never set, and your attestation is on the report. This is a plain look at where AI genuinely helps with survey write-ups, where it does not, and a set of prompts you can paste in and test on your own real notes.
The honest picture
AI is genuinely useful for turning the work you already did on the boat into prose:
- What AI does well today: take your terse field notes and photo labels and write each finding in clear, consistent language, hold a steady tone a bank or underwriter will read easily, apply the condition ratings you already assigned, and draft the scope, summary, and recommendations sections from what you recorded. It compresses the report-writing hours, not the survey.
- What AI does not do: decide what is a deficiency, what condition rating a component gets, or what the vessel is worth. It cannot stand in the bilge, sound the hull, or run a sea trial, and it cannot sign the report. The report is issued under your name and your attestation, and you are accountable for every finding, rating, and value in it. AI can phrase a finding you made; it must never invent one, change a rating you set, or assert a value you did not provide.
The right way to think about it: AI is a fast writer working from your notes, not a second surveyor. The observations, the ratings, and the valuation are yours. The typing is what you hand off.
The line: it will invent findings and values if you let it
The specific failure to watch for is fabrication, and it shows up three ways:
- From a thin note. Give AI "stbd chainplate, surface corrosion" and it may write a confident paragraph about rigging failure risk and a recommendation to re-rig that you never observed or wrote. It fills gaps with plausible-sounding detail. In a survey report, plausible-sounding detail you did not observe is a finding you did not make.
- From a photo. It is tempting to point a vision model at your photos and ask "what is wrong here." That is the exact line not to cross. A model guessing a defect from a picture is inventing findings, and the bank and the underwriter rely on the report as if an accredited surveyor observed it firsthand. Use the photo to organize and label, never to diagnose.
- From the particulars. Hand it the make, model, and year and it will happily produce a fair-market or replacement value to "round out" the report. That number is yours to set from the market and the vessel's actual condition. A value the model guessed is a value you did not stand behind, and a lender or insurer will act on it.
The fix is the same in every case: AI describes only what your note already says, at the rating you already set, with no value you did not provide, and flags anything it cannot tie to one of your observations rather than filling it in.
The setup that keeps the report yours
Two habits make AI much safer for report writing, and the prompts below build them in:
- Give it your findings, your ratings, and your numbers. Hand it your notes and the exact condition rating scale you use, and your valuation if the report includes one. Tell it to use only those, to never upgrade or downgrade a rating, and to assert no value you did not write down.
- Make it flag, not fill. Tell it that if a note is too thin to write a finding from, it must say so and ask, not invent the rest. You want a draft that surfaces your gaps, not one that papers over them.
How to test it on your own work
Do not trust a polished demo, including this one. Pull two or three of your own recent surveys, the raw notes and photo labels, a timer, and the prompts below. Rate each output 1 to 5 on usefulness and accuracy, and compare the time against how you write the report today. Keep what wins.
Paste-ready prompts
Copy these as written. Bracketed text is what you swap per survey.
Test 1: Write the findings from your notes and photo labels (text model)
I am giving you my raw field notes and photo labels from one marine survey,
plus the exact condition rating scale I use. Turn each note into a clear
written finding. Rules:
- Write only what my note says. Do not add a cause, a consequence, a location,
or a condition I did not write down. Do not diagnose from a photo label,
treat the label as a caption only.
- Use my condition ratings exactly as I assigned them and never change them:
[paste your rating scale, for example new / good / fair / poor / not examined].
- Assert no value, age, or specification I did not provide.
- If a note is too thin to write a finding, do not invent the rest: list it
under "Needs my input" and ask me what to add.
Rating scale: [paste your scale]
Field notes and photo labels: [paste them]
Watch for: did it stick to your observations, or did it write a paragraph you never said? Every invented detail and every changed rating is the work it cannot do for you.
Test 2: Write the recommendations section (text model)
Here are the deficiencies I listed on this survey, each with my condition
rating: [paste them]. Write the recommendations section using only these
deficiencies. For each, state the recommended action in plain language.
- Add no recommendation for anything not in the list.
- Do not raise or lower any condition rating I assigned.
- Do not assert a repair cost or a value. If urgency matters, use only the
priority wording I provide: [paste your priority terms if any].
Watch for: does every recommendation trace back to a deficiency you actually listed, and did it stay off cost and value entirely?
Test 3: Draft the summary and scope from the particulars (text model)
Here are the vessel particulars and the scope of this survey as I recorded
them: [paste them]. Draft the summary and scope-of-survey sections.
- Use only the particulars and scope I gave you. Add no finding, no condition
assessment, and no value of any kind.
- Do not state the vessel's worth, market value, or replacement cost. I set
those separately, leave them out.
- If a particular is missing, mark it [TBD] rather than guessing it.
Vessel particulars and scope: [paste them]
Watch for: did it leave value out completely, and did it mark missing particulars instead of filling them with a plausible guess?
Test 4: Fabrication and consistency audit (text model)
Review this draft survey report for internal problems only. For each issue,
quote the exact line and say what is wrong.
1. Findings not tied to a note: any statement that reads like an observation or
conclusion with no basis in my field notes.
2. Ratings I did not assign: any condition rating that does not match the one
in my notes for that component, or any rating applied where I gave none.
3. Values I did not set: any fair-market value, replacement cost, repair
estimate, or dollar figure I did not provide.
Do not add findings, do not change my ratings, and do not supply any value.
Only flag.
Field notes: [paste them]
Draft report: [paste it]
Watch for: does it catch the finding that has no note behind it, the rating that drifted, and any number that crept in? Run it on a report you already delivered.
What success looks like
If your own testing shows real time savings, the next step is a small pilot: run a week of surveys through the prompts and measure the hours. If that holds up, the natural next step is a simple agent, running on your own cloud, that you use in plain language. The most useful version takes your structured field notes and photo labels, drafts each section using only your condition ratings and the values you provide, and cites the exact note behind every finding, surfacing anything it cannot tie to one of your observations, so you verify and sign in minutes instead of hours.
The principle holds the whole way through: AI gives you a faster draft and a second set of eyes on consistency. It does not stand in the bilge, it does not set the value, and it does not sign the report. Keep that line clear and the rest is upside.
This is a workflow guide, not a substitute for a licensed or accredited marine survey. Every finding, condition rating, valuation, and the attestation on the report are the surveyor's own professional judgment and responsibility.
Want a straight answer for your survey business?
I build practical AI and custom software for businesses, on Google Cloud. If you want a second set of eyes on how AI could fit your reporting workflow, or on a tool you are considering buying, tell me what you are working with. No pitch, just a straight answer.