GUIDES
Find The Money Hiding In Your Inbox 📚
Every unused credit, refund and gift card you forgot about. I ran this on my own inbox and found 800 dollars of airline credit. Takes about 20 minutes.
I ran this on my own email expecting a stray twenty dollar gift card, and it came back with 800 dollars of American Airlines credit that doesn't expire until 2028. I had no memory of it. If I had never gone looking I would have let the whole thing quietly disappear.
It also turned up a handful of restaurant gift cards I'd been given and forgotten about, which is somehow worse, because those were presents from people who spent real money on them.
Why This Works So Well
43% of Americans are holding at least one gift card, voucher or store credit they've never used, and the average person in that group is sitting on 244 dollars of it. Across the country there's north of 20 billion dollars in gift cards nobody ever spent.
Every company that hands you a credit is quietly hoping you forget about it. When an airline gives you a flight credit instead of your money back, they keep your cash and they know a slice of those credits will never get used. None of it is a scam, it's just how the maths works. They're counting on the fact that a credit is a boring email you got eighteen months ago, buried under four thousand others.
That's exactly the problem AI is good at. Your inbox already has every receipt, every confirmation, every apology email with a code in it. You were never going to read back through nine years of email by hand. Something else can do it in two minutes.
Step 1: Connect Your Inbox
Nothing works until the AI can read your email. This is the only setup step and you only do it once.
If you use Claude (recommended, works on the free plan):
- Click your name or profile at the bottom left, then Customize. You can also go straight to claude.ai/customize/connectors.
- Open Connectors, then click the + button or Browse connectors.
- Search Gmail, click it, then click Connect.
- Google shows its own sign-in screen. Pick the account you actually want searched, then approve.
If you use ChatGPT:
ChatGPT calls these Plugins now. Click Plugins in the left sidebar, or Settings then Plugins, or go to chatgpt.com/plugins. Search Gmail, click Connect, sign in to Google, approve.
What it can and can't do: it reads. That's it. It cannot send an email as you, delete anything, or change a single thing in your inbox. Claude can write a draft if you ask, and that draft just sits in your drafts folder waiting for you to press send.
To cut it off: Claude, go to Customize then Connectors then Disconnect. ChatGPT, go to Plugins and disconnect. Or from Google's side at myaccount.google.com/connections, which works no matter what the app says.
Step 2: The Main Prompt
This is the whole guide in one paste. Open a new chat, paste it in, and let it run. It takes a couple of minutes because it's genuinely reading back through years of your email, so don't close the tab the second it looks like nothing is happening.
Search my entire Gmail history for any money I am owed
or have never used. Include gift cards, e-gift cards,
store credit, airline and travel credits, vouchers,
refunds that were promised, deposits, rebates,
settlement payouts and unused promo credit on any
account. For each one give me: the company, the amount,
the date of the email, the code or reference number,
the expiry date if there is one, and a direct link to
the original email. Sort by highest value first and
add up the total at the bottom. Do not invent anything.
If you are unsure about an amount, say so.
Step 3: The Validation Prompt
A list of credits from your email is not the same thing as money. Some codes are already spent, some quietly expired, and some were never worth what the email said.
This one needs a browser. With Claude, that means the Claude Chrome extension connected. With ChatGPT, use the desktop app, switch to Work, and type @browser in your message.
Now go to each of these companies' websites and
actually check whether each credit or gift card is
still valid and how much is left on it. I'll log
in wherever you need me to. Come back with three
lists: confirmed still good with the real balance,
expired or already used, and the ones you couldn't
verify.
You'll have to help it here, and that's normal. It'll get to an airline site, hit a login screen, and stop. Log in yourself, then tell it to carry on. It will not ask for your passwords and you should never type one into the chat.
The middle list is the one that saves you a wasted afternoon. Knowing a credit is dead is worth almost as much as knowing one is alive. The third list is your own to-do list of two or three phone calls worth making.
Four More Prompts Worth Running
Run these in the same chat so it still remembers everything it found.
Put everything you found into one clean table sorted
by expiry date, with the company, amount, code,
expiry and a link to the email. Flag anything
expiring in the next 90 days in a separate list at
the top.
Set out a plan for using the ones expiring in the
next 90 days, what to use first and what I'd spend
it on.
Find every subscription I'm paying for that I haven't
used in six months, with the cost, the renewal date,
and what I'd save in a year.
Find any refund I was promised in an email but never
actually received. Show me the original email, the
amount, the date it was promised, and whether
anything later confirms it arrived.
A company says "your refund has been processed and will appear in five to seven business days," and then sometimes it just doesn't, and nobody is ever going to chase it except you.
Bonus, because it's the same idea
Search my Gmail for any class action settlement
notices, recall notices, or emails saying I might be
eligible for a payout or compensation. For each one,
tell me the deadline to file and whether I still
have time.
These land in your inbox looking exactly like junk, get ignored, and often carry a real payout with a hard deadline attached.
Six Other Places The Money Hides
Your inbox is the biggest single pile, but it isn't the only one. Check these by hand.
Airline accounts. Log in to each airline you've flown and look for a credits, vouchers or wallet section. Airlines are notorious for holding a credit in your account without ever emailing you about it again.
Hotel and travel loyalty. Point balances, and whether they expire from inactivity.
Store accounts. Store credit from returns sitting on the account rather than on a card.
PayPal and Venmo. Money someone sent you that you never moved to your bank.
Physical cards in a drawer. Anything that never touched your email will never show up in this search.
Unclaimed property. The one almost nobody knows about. Every US state holds money that belongs to you: old deposits, final paychecks, forgotten accounts. Search your name for free at missingmoney.com and on your state's own programme.
The Two Rules
It only sees what's in your email. Anything that lived only in an app, or on a physical card in a drawer, will never show up. Check those separately.
Verify before you count it. Codes get spent, expire, or come with terms that changed. It isn't money until you've seen a live balance on the company's own site.
Most people find something between fifty and a few hundred dollars, and a decent number find nothing at all, which is a perfectly fine outcome and means your record keeping is better than mine. It costs you twenty minutes once and the downside is zero.
Your Action Plan
- Connect your inbox. Use whichever account holds your older receipts.
- Run the main prompt. Let it work through years of email.
- Validate everything with the browser prompt before you count a single dollar.
- Diary every expiry date. Put each one in your phone calendar with a reminder two weeks before, and a note a year from now to run the whole thing again.
It's not found money. It was always yours. You just forgot where you put it.
The real unlock of AI agents is not full autonomy, but human-in-the-loop systems design. Structure your agent inputs with explicit constraints, negative prompts, and automated test checkpoints. When building tools, keep token consumption lean by caching system prompts and isolating tasks into specialized sub-agents.