The Amazon email worth $51 you probably deleted (and 5 more like it)

Today's Tip: Finding The Class-Action Money That's Already Yours

You almost certainly have unclaimed settlement money with your name on it right now. And nobody is going to send you a letter about it.

Let me explain…

Last month my friend Kathy got a text from her son David. It was a link to an Amazon Prime settlement notice, and the deadline to file a claim was three weeks away.

Kathy is 66, retired eight years ago from a dental hygienist career in Duluth, Minnesota. She had been an Amazon Prime member since 2019. But she had never heard of the settlement, and she had never received the notice email. It had landed in her promotions folder six months earlier and sat there, unopened. The Amazon settlement alone caps individual payouts at $51… small change, until you realize it is just one of dozens of open cases in any given year.

Here is what surprised her most… when she started looking, the Amazon case was not the only one she qualified for. Not even close.

She had bought bagged salad in 2019. She had had a Comcast account in 2022. She had owned a Kia Sorento. She had used Fandango to buy movie tickets. Every one of those was tied to an open class-action settlement fund, sitting on a court-ordered pile of money, waiting for people like her to raise their hand.

She used one AI prompt to find them all. Filed six claims in about 45 minutes at her kitchen table. And she expects somewhere between $180 and $600 in checks to arrive over the next year… all from claims she would never have known existed.

Here is the prompt she used.

★ Copy-Paste Prompt

Works in Claude, Gemini & Grok

Act as a class-action settlement research assistant. I am going to give you a list of products, services, and purchases I have used in the past 10 years. Your job is to find every open class-action settlement currently accepting claims in the United States that I might qualify for.

Here is my list:
- Amazon Prime member from [year] to [year]
- Cell phone carrier: [carrier and years]
- Bank or credit union accounts: [names and years]
- Cars I owned or leased: [make, model, year]
- Streaming services I subscribed to: [list]
- Any data breach notices I received: [company names]
- Grocery stores I regularly shopped at: [names and state]

For each open settlement I might qualify for, give me:
1. The official case name
2. The eligibility criteria (who exactly qualifies)
3. Whether proof of purchase is required
4. The estimated payout range
5. The claim deadline
6. The official claim form URL — it must be a court-approved administrator site, not a third-party aggregator

At the end, tell me which one I should file first based on deadline. Be conservative. If you are not sure I qualify, say so.

Why this works: You are giving the AI a shopping list of your consumer history and asking it to match against currently-open court cases. The instruction to cite only the official administrator URL (not a third-party aggregator) is the piece that keeps you off scam sites. And asking for the deadline first tells you which claims are urgent versus which can wait until next month.

Two things make this prompt work well with Grok in particular. First, Grok pulls from the live web… so it knows which settlements are still open today, not six months ago. Second, when it gives you a link, it is a link to a page it has actually seen in the last few days, which means the claim form usually still works.

When Kathy ran this prompt, she got back a list of eleven possible matches. She dismissed five as too weak (she could not remember for certain whether she had bought a certain brand of eye drops). She filed six. Each claim form took her under five minutes to complete.

★ Today's "Aha!" Moment

A class-action settlement is not a lawsuit you have to join. It is a court order that already exists. A judge has already told the company to pay. The money is already sitting in a settlement fund. Your only job is to raise your hand and say, "Yes, that was me."

Now, I know what you are thinking… "This sounds like the sketchy emails I already ignore." I understand. Those emails do exist. The FTC gets thousands of complaints a year about fake settlement notices that ask for your Social Security number or a "filing fee."

But the difference is this… when you look up an open settlement yourself, using an AI tool to check current cases, you skip the email trap entirely. You go straight to the official claim form. No one asks you for money. No one asks for your bank login. Most claim forms need only your name, address, email, and a checkbox saying which product you used.

★ One Thing to Watch

Never file a claim through a site that asks for a filing fee, a bank login, or your full Social Security number. Legitimate settlement claim forms are free to submit and are hosted on court-approved administrator sites. If a link Grok gives you redirects to something asking for payment, close the tab. Grok is helpful… but it is still your job to double-check every URL before you type anything into it.

One afternoon. Six claims. Done for the year. Kathy set a reminder on her phone to run the same prompt every six months, because new settlements open constantly and old ones close.

You almost certainly have money waiting. Go find it.

★ ★ ★

Tool Spotlight ★

Grok

Site: grok.com

Made by: xAI

Cost: Free tier available

When your question depends on today's answer instead of last year's, Grok is the one you reach for. Its strength is the live web… every response is stitched together from pages it has just visited, which means it knows what deadlines are open, what recalls are active, and what a product costs at the store right now. For a task like finding open class-action settlements, that is the whole ballgame.

The interface is plain by design. You type a question, it answers, and every important claim usually arrives with a source link you can click to verify. That last part matters. Grok is not going to invent a settlement that does not exist, but if it ever slips up, the missing source link is what makes the slip obvious.

Best for: Current events research, product recall lookups, live price comparisons, verifying whether a news story is real, and any question where the correct answer depends on what happened yesterday.

★ ★ ★

★ Quick Wins

Check your closet for active recalls. Ask Grok: List every active product recall in the past 18 months for [product category — for example: cordless power tools, kitchen appliances, or space heaters]. Include the model numbers affected and how to get a refund or free replacement. This is powerful for the things you have owned a while and forgotten about. Works best in Grok because it pulls the current CPSC recall database live.

Get today's cash price for your prescription. Type into Grok: What is the current cash price for a 30-day supply of [medication name and dose] at CVS, Walgreens, Costco, and Sam's Club pharmacies in the [your ZIP code] area? Also compare to the GoodRx and Mark Cuban Cost Plus Drugs price. The gap between the highest and lowest is often $40 to $100 a month. Works best in Grok for the live pricing.

Find free events near you this weekend. Ask Grok: List every free or low-cost community event happening in [your city and state] this Saturday and Sunday. Include events at libraries, museums, community centers, and public parks. Note which ones are family-friendly for visiting grandchildren. Great for planning grandkid visits… without spending anything. Works in Grok and Gemini.

Skill Builder: The Self-Audit Prompt

Here is a habit that will make you a sharper AI user in about 30 seconds… ask the AI to audit itself. Every important answer should be followed by a second question — "Now check your work."

This is different from what we covered on Tuesday, which was asking AI to argue against its own recommendation. That is the Devil's Advocate move. The self-audit is a smaller step… you are not asking for a different answer. You are asking the AI to flag its own weak spots before you act on what it told you.

The trick is being specific. "Are you sure?" is a poor follow-up. AI models are trained to sound confident, so if you ask them vaguely, they will usually just repeat themselves in a slightly different tone. What actually works is asking for three specific things — what could be wrong, what got left out, and what you might not have known to ask.

I use this after every AI answer that involves money, health, or a decision I am about to act on. It takes 20 seconds. About one time in five, the AI catches something meaningful — a stale statistic, an assumption I did not tell it about, a step I had not considered. The other four times, it confirms the original answer, which is also useful. Either way you save time.

★ The Self-Audit

Works in Claude, Gemini & Grok

Before I act on your last answer, run a self-audit.

1. List three specific ways your response could be wrong or incomplete.
2. List two important questions I probably should have asked but did not.
3. Rate your confidence in the original answer on a scale of 1 to 10, and tell me exactly why.

Be honest. If you were guessing on any part, say so plainly.

Why this works: You are forcing the AI out of "here is a confident answer" mode and into review mode. The specific numbers — three ways, two questions, 1 to 10 — prevent it from giving you a vague "well, of course anything could be wrong" non-answer. Naming the parts you want it to check separately is what makes the review actually sharp.

Try it on any AI answer this week where you were about to make a decision that matters — a purchase, a call to the doctor, a letter you were about to send. You will be surprised how often the self-audit surfaces one thing worth changing.

Fair warning… you will start doing this with humans too. It is a hard habit to un-learn.

The Amazon email worth $51 you probably deleted (and 5 more like it)

Read More from the AI Advantage