My first job out of university was in the marketing department at American Express, and I spent seven years there marketing exactly this kind of thing.

Here is what that teaches you. The benefits people miss are almost never the automatic ones. Those take care of themselves. The ones people miss are the ones you have to switch on. The lounge access that needs you to enrol in the lounge program separately. The credit that only pays out if you registered the card first. The thing that quietly resets on the 31st of December whether you used it or not.

Nobody is hiding these from you. They are usually sitting on a page you have never opened, written in a way that assumes you will go and do something about it. Most people never do, because there is no single place that says here is your list, here is what you have to do, here is when.

That is the list you are about to build. Five prompts, in order, in one chat. Any AI you already have. The free version is fine.

What you need before you start. Your cards in front of you, or your statements open. You need the exact card name including the tier, because "Gold" and "Gold Preferred" are different products with different benefits. And you need to know which country your cards were issued in. That second one matters more than anything else in this guide, and I explain why in step one.

What this guide will not do, on purpose. It will not tell you which card to get, compare cards, or work out whether an annual fee is worth it. That is financial advice in a regulated category and I do not do that. This organises what you are already paying for. Every prompt below is written to stay on the right side of that line, and if your AI volunteers a recommendation anyway, ignore it.

Step 1: Make it interview you

The instinct is to type "what are the benefits on my Amex Gold" and see what comes back. Do not do that. You will get a confident, tidy, partly wrong answer, and you will have no way of telling which parts are the wrong ones.

Start by making it collect the facts first, one card at a time, and confirm them back to you.

Prompt 1 of 5: the intake

I want to build a complete picture of the credit card benefits I already have and am not using. Do not list any benefits yet. We are only gathering facts in this step. First, ask me which country my cards were issued in, and confirm it back to me before you go on. Then ask me about my cards one at a time. For each card ask me for: the exact name as it is printed on the card or as it appears on my statement, including the tier or version; who the issuer is, which may be a bank rather than the card network; and roughly what year I opened it, if I remember. Wait for my answer before you ask about the next card. When I tell you that is all of them, repeat the full list back to me with the country, and ask me to confirm it is right before you go any further.

The country question is not filler, and it is the single most common reason these lists come out wrong. A card with the same name in the United States and in Canada is not the same product. Different credits, different partners, different terms, sometimes a different insurance provider underneath. I am Canadian and most of my audience is American, and I have watched this exact thing send people looking for a benefit that does not exist where they live.

The year you opened it matters for a smaller but real reason: benefits get added and withdrawn, and some issuers grandfather older accounts onto terms that new applicants do not get. If you cannot remember, say so and move on. It is a nice to have, not a blocker.

Step 2: Make it verify, and make it admit what it could not

This is the step that makes the difference between a useful list and a list that embarrasses you at an airport counter. An AI listing card benefits from memory will get some of them wrong. Not because it is bad, but because amounts change, partners change, and it learned the 2023 version of your card.

So you make checking part of the job, and you make not knowing an acceptable answer.

Prompt 2 of 5: the verify step

Now, for each card on that list, find the current benefits for my country using the card issuer's own website. Not blogs, not forums, not points or travel sites, not comparison sites, and not your own memory of what these cards used to include. For every benefit you give me, tell me which page you took it from and how confident you are that it is current and correct for my country. Then give me a separate list headed "I could not verify these" and put in it anything you think may exist but could not confirm on the issuer's own pages for my country. That includes anything where different sources disagreed, anything where the amount or the terms looked different depending on where you looked, and anything you are working from memory on. Do not quietly leave those out and do not fill in the numbers with your best guess. A short verified list plus an honest list of unknowns is far more useful to me than a long list I cannot trust.

Treat that second list as your to do list for the issuer's app, not as a problem with the prompt. It is doing what you asked.

The higher accuracy version, if you want it

There is a way to make this close to airtight, and it costs you about two minutes per card. Instead of asking the AI to go and find the benefits, you bring the benefits to it.

Open your issuer's benefits page for your card, select the whole page, copy it, and paste it in. Now the AI is organising real text rather than recalling what it thinks the text says. That is a completely different task and it is much harder to get wrong.

Prompt 2, paste in variant

Below is the benefits page for my [card name] from [issuer], for [country], copied directly from their website today. Use only this text. Ignore anything you already believe about this card, and if the text below contradicts what you thought you knew, the text below wins. [paste the page here] Read it and tell me, using only what is in that text: what is automatic, what I have to switch on or enrol in, and what expires or resets on a date. Quote the wording that made you put each item where you put it. If the page is genuinely unclear about something, say it is unclear rather than deciding for me.

Do that for each card and paste each one in turn. If you only do it for one card, do it for the card you pay the highest fee on.

Step 3: The three lists

Now you sort it. This is the part that makes the whole thing usable, because the three lists are three completely different jobs. One is information, one is a task, and one is a deadline.

Prompt 3 of 5: sort it into three lists

Now take everything you verified and split it into three lists. Every benefit goes in exactly one list. 1. AUTOMATIC. Things I already have without doing anything at all. I do not need to act on these. I only need to know they exist so I stop paying separately for something I already have. 2. SWITCH ON. Things that only work if I enrol, register, activate, opt in, or add my card to a program. This is the list that matters most, so put it first in your reply. 3. EXPIRES. Things that reset or disappear on a date, whether that is monthly, quarterly, annually, or on my card anniversary. For every item give me: which card it belongs to, one plain sentence on what it actually gets me in practice, and for anything in list 2 or list 3, exactly where I go to do it, whether that is a page in the app, a phone number, or a website. If something belongs in more than one list, put it in the more urgent one and tell me why. Keep the unverified items separate at the bottom under their own heading so I do not confuse them with the rest.

Most people are surprised twice here. Once at how much is sitting in list 1 that they had no idea about, and once at how short list 2 is. That second one is the good news. The switch on list is usually a handful of things, and it is almost always an afternoon of work at most.

Which brings us to the two prompts that are not in the video.

Step 4: Turn the list into one sitting

A list is not the point. I have watched people build a beautiful list of benefits and then never switch a single one on, because "enrol in the lounge program" is sitting in a note somewhere with no time attached to it and no idea whether it takes four minutes or an hour.

So this prompt does two things. It turns list 2 into a single sitting you can actually book, and it turns list 3 into a calendar for the rest of the year.

Prompt 4 of 5: the one sitting sprint, and the year calendar

Take those three lists. I want two things from them. First, the sprint. Give me every benefit I can switch on once and then never think about again: enrolments, registrations, activations, adding a card to a program. Order them shortest task first. For each one tell me roughly how long it takes and exactly where I do it. Then give me a total time for the whole list, so I can sit down and do the entire thing in one go rather than one at a time over six months. Second, the calendar. Give me a month by month calendar for the rest of this year of every benefit that has a deadline or a reset date. For each one tell me what it is, exactly what I have to do to claim it, the real deadline, and the date I should actually act on it, which is three weeks before the deadline and not the day of. If a benefit resets monthly or quarterly, show me the next two reset dates rather than listing every single one for the year. Do not recommend any cards, compare my cards to other cards, or work out whether any annual fee is worth paying. That is not what I am asking for.

The three weeks instruction is deliberate and it is the part people cut. A reminder on the deadline is not a reminder, it is a notification that you missed something. Three weeks is enough time to actually book the thing, order the thing, or sit on hold with the issuer.

Now look at the total at the bottom of the sprint. For most people it is somewhere under an hour. That is the entire gap between paying for these benefits and having them.

Step 5: The reminders, and the version that renews itself

Last one, and it is the one that decides whether any of this still exists in March.

Prompt 5 of 5: the reminders, and the January re-run

Now write the exact calendar entries I should create on my phone, ready to copy in. For each one give me: A title I will still understand at a glance in six months, with the card name in it, not a vague one like "card benefit". The date, which is three weeks before the deadline. One line telling future me exactly what to do and where to do it, written so that I never have to come back and re-read any of this to act on it. Keep each entry short enough to fit on a phone notification. Then, separately, write me one prompt I can paste back to you every January to redo this entire exercise from scratch. It should assume I am starting a new chat with no memory of this one, ask me for my cards and my country again, run the same verification step, and produce the same three lists, the same sprint and the same calendar for the new year. These benefits change every year, so the list we just built will be out of date by then and I want to be able to rebuild it in one paste rather than remembering how we did it.

That last paragraph is the most valuable thing in this guide and it is the part that gets skipped, so I will say plainly what it does. It turns a one off afternoon into something that runs itself every year. Card benefits genuinely change annually. Credits get added, partners get dropped, amounts move. A list built today is a good list in December and a misleading one next June.

Save the January prompt somewhere you will find it. In a note, in your saved prompts, pinned in the chat, wherever you actually look. Then put one more calendar entry in for the first week of January that says "paste the card benefits prompt" and you are done thinking about this forever.

What to actually do this week

If you read this far and do nothing, the guide failed. So, smallest possible version:

  1. Today. Run prompts 1, 2 and 3 on your highest fee card only. Not all your cards. One. It takes about ten minutes and it will tell you whether this is worth the afternoon.
  2. This week. Book the sitting. Look at the total time prompt 4 gives you, find that slot in your week, and do the whole switch on list in one go with the app open next to you.
  3. Before you close the tab. Put the reminders in your phone from prompt 5, and put the January entry in. This is the ten minutes that makes the other hour permanent.

And when something on the "could not verify" list turns out to be real, check it against your issuer's own page rather than the chat. The AI got you the list. Your issuer is still the only thing that gets to be right about what you have.

Sources & Notes

  1. This guide does not recommend, compare, or rank credit cards, and it never will. No card suggestions, no annual fee maths, no rewards optimisation. It organises benefits on cards you already hold. Anything beyond that is financial advice in a regulated category and I am not a licensed advisor. If your AI offers a recommendation while you are running these prompts, that is the AI going off script.
  2. The issuer's own page is the only authority here. Every prompt above is written to send you back to it, and prompt 2 exists entirely for that reason. Nothing in this guide states what any specific card includes, because that is not something a guide can be right about for long.
  3. Why the country question comes first. Cards sold under the same name in different countries are different products, with different credits, different partners and frequently different insurance underwriters. This is the most common way an AI generated benefits list goes wrong, and it goes wrong invisibly, because the answer looks completely reasonable.
  4. Why the card tier and the year matter. Tiers of the same card carry different benefits, and some issuers keep older accounts on terms that are no longer offered to new applicants. If you cannot remember the year, skip it. It sharpens the answer, it does not gate it.
  5. On AI accuracy with card benefits specifically. Benefit amounts, partners and terms change frequently, which makes them exactly the kind of fact a language model recalls confidently and wrongly. That is why verification is built into the method as a step rather than added at the end as a disclaimer, and why the paste in variant in step 2 is the most accurate version of this whole exercise.
  6. Any AI works for this. There is nothing here that needs a paid plan, a specific tool, or any feature beyond a normal chat window. If a step only works on a paid tier, it does not belong in this guide.
  7. The American Express reference is biographical. I worked in the marketing department there for seven years, starting straight out of university. Everything in this guide about how card benefits work is a consumer side observation that anyone could make from the outside. I do not speak for American Express, I have no current inside knowledge of their products, and nothing here is drawn from anything internal.