Can ChatGPT Make a Home Inventory for Insurance?

Short answer: it will produce a list, and a reasonable one.

Point a capable AI assistant at photos of your living room and it will tell you there is a three-seat sofa, a large television, a coffee table and a rug. It will guess at brands where they are visible and estimate what things cost. We are not going to pretend otherwise, and you should be sceptical of anyone who does, because you can test it yourself in about a minute.

But answering "can it" is not the same as answering "should you", and there are two separate reasons the answer to the second one is no. The first is that you will do noticeably more work for a noticeably worse record. The second, and the one that actually costs you money, is that when you are finished you will be holding something that does not count as evidence, and cannot be made to count no matter how good the AI gets.

Take the second one first, because it is the one nobody tells you.

The list is not the hard part

Here is the thing most people get wrong about home inventories. They assume the difficult bit is the work: the walking around, the remembering, the pricing, the typing. So a tool that removes the work feels like it has solved the problem.

It has solved a problem. It has not solved the one that costs you money.

When you make a contents claim, an assessor is not sitting there wondering whether your list is neat, or whether your values look about right. They are asking two much harder questions:

1. Did this record exist before the loss?
2. Has it been changed since?

Everything else is detail. Those two questions are the whole game, because the answer to them determines how much weight your document carries, and the weight of your document determines your payout.

Why a self-made document cannot answer them

A list you generated yourself cannot answer either question. Not because you are dishonest, and not because the AI did poor work. Because of who made it.

Its creation date is whatever your computer says it is. Its contents can be edited freely, at any time, before or after the event. And crucially, you are the person who benefits from it. A claims process is built on the assumption that the claimant is an interested party. Self-attestation is precisely what it is designed to test.

That is not a flaw in ChatGPT. It is a structural limit on anything you produce for yourself, and it applies equally to a spreadsheet, a Word document, or a beautiful AI-generated PDF. No amount of model improvement will fix it, because the problem was never intelligence.

What actually goes wrong

It is worth being precise about the failure mode here, because people imagine the wrong one.

The risk is not that an adjuster accuses you of fraud. That is rare, and it is not what happens.

What happens is quieter and far more common. Your unverifiable list simply does not carry much weight, so it does not move the number. The assessment drifts back toward what the adjuster is willing to assume about a home of your size, which is almost always substantially less than what you actually lost. You are not disbelieved. You are discounted.

That is how underinsurance turns into underpayment: not through an accusation, but through an absence.

What does count as evidence

Look at what assessors actually give weight to, and a pattern appears immediately:

• Dated photographs and video of items in your home
• Receipts and bank or card statements
• Serial numbers
• Professional valuations
• Anything held by a third party

Every single one has the same property. It comes from somewhere other than you, or it can be checked without trusting you. That is the entire definition of evidence, and it is exactly the property an AI-generated list lacks.

You cannot notarise yourself

This is the sentence worth taking away.

A signature is only worth something because the person who benefits from the document could not have forged it. The moment you could have produced it yourself, it stops proving anything. Which means proof, by its nature, can never be a thing you generate for yourself, no matter how good the tool.

This is why notaries exist. It is why witnesses exist. It is why a receipt from a shop is worth more than a note you wrote saying you bought something. Not because you are untrustworthy, but because "trust me" is not a mechanism.

The other reason: it is more work, for less

Everything above is about why an AI-made list will not stand up. But there is a plainer objection too, and it is worth saying because people assume the chatbot route is the lazy one.

It is not. You have to photograph every room yourself, upload the images, prompt it space by space, keep it from inventing things, chase it for the items it skipped, and then assemble the output into something coherent. It has no idea what a home inventory is for, so it will not ask about the serial number on your laptop, or notice that a ring worth four thousand dollars deserves different treatment from a toaster.

And you do not need to do any of that, because the inventory is the part WHIG gives away.

You talk through your home once, out loud, the way you would show it to a friend. WHIG pulls the items, the brands and the values from what you said and what it saw, organises them by room, lifts the frames from your own footage as visual proof, and flags what needs strengthening. That costs nothing. No card, no trial.

So the honest comparison is not a free chatbot against a paid app. It is a free chatbot list against a free WHIG inventory that is better, faster, purpose-built, and can be sealed into evidence.

Where that leaves you

An AI assistant will hand you an unsigned, undated document that you produced yourself, which is the exact category of evidence a contents claim is built to discount. You will have worked harder for it than you expected.

The part the AI cannot touch, at any price and at any level of capability, is a record sealed and timestamped at the moment you make it, by something you do not control, so that a stranger can confirm it is genuine without taking your word for anything.

That is what WHIG adds. Your walkthrough is hashed and cryptographically signed with a key held in a hardware security module that we never export and you never hold. The record is timestamped the moment it is created and tamper-evident from then on. Anyone you hand it to can verify it themselves, in their own browser, with nothing uploaded, at whig.app/verify. Change one character and the check fails.

Record your home free, see the whole inventory free, and judge the work for yourself before you spend anything.

If you never intend to claim, a list is fine.

If you might, a list is not what you need.

Estimates produced by WHIG are not professional valuations. Specialist and high-value items should be assessed by a qualified valuer.