How Can I Check Whether AI Assistants Invent Coupon Codes for My Online Store?

Why does it matter?

• Checkout frustration. A shopper who tries a code that fails may abandon the basket, or blame your store.
• Support load. "The code ChatGPT gave me doesn't work" becomes a regular ticket.
• Margin leakage. A leaked partner or staff code that still works gives discounts to people it was never meant for.
• Distorted partner reporting. Influencer and affiliate codes used by the general public inflate those partners' apparent sales.
• Trust. Repeated invalid codes make an assistant's answers about your store look unreliable, and some shoppers conclude the store is a scam.

Which questions should you test?

• "[Brand] discount code"
• "[Brand] promo code [current month]"
• "Is there a [brand] first-order discount?"
• "[Brand] student or military discount"
• "[Brand] free shipping code"
• "Best way to save at [brand]"

Ask them in ChatGPT, Gemini, Perplexity, Copilot and Google's AI features, three to five times each, with web search on and off. Record the exact codes, the conditions stated, and any cited sources.

How do you classify the codes?

Also check what the answer says about conditions, such as minimum spend, eligible products and expiry. Wrong conditions produce the same frustration as wrong codes.

Where do invented codes come from?

• Coupon aggregator sites that list many codes per store, including expired ones and guesses, and that rank well for "[brand] promo code" searches.
• Leaked partner codes shared beyond their audience.
• Old promotions still mentioned in blog posts, newsletters and social posts.
• Pattern guessing. Assistants may generate a code that looks like a typical discount code, especially when no reliable source exists.
• Browser extensions and deal forums that circulate codes.

Open the sources the assistant cites. Invented codes often trace back to one or two aggregator pages, so fixing those sources can clear several wrong answers at once.

What should you publish?

Create one official offers page, linked from your footer and help centre, that says:

• Current public offers and codes, with conditions and end dates.
• Where genuine codes are shared, such as your newsletter, account area and named partners.
• What does not exist: "We do not offer a general discount code beyond those listed here." This gives assistants a clear statement to repeat.
• Standing offers, such as student, military or first-order discounts, with how to claim them.

Keep the page current. An offers page that lists an expired code becomes another source of wrong answers. How to mark discounted prices so assistants report them correctly is covered in how subscription, discount and coupon prices should be marked so AI reports the right price.

What else should you change?

• Retire leaked codes, and give partners unique codes, or single-use codes where your platform allows.
• Improve the invalid-code message at checkout: "This code isn't valid. See our current offers" with a link. A dead end becomes a second chance.
• Ask aggregators to remove fake and expired codes. Many have a contact or correction form, and some accept merchant-verified listings.
• Report answers with invented codes through each assistant's feedback options.
• Watch support tickets for mentions of codes from AI assistants, and add the codes to your test list.

How do you track progress?

Keep a simple log of each round: the questions, the codes given, their classification and the sources. Track the share of answers that give only valid, correctly described codes. Rerun the test monthly and after each promotion ends, since that is when a newly expired code starts giving shoppers failed checkouts.

What might an illustrative case look like?

Consider an illustrative skincare store whose support team received a dozen emails a week about failed codes. Testing showed assistants offering "GLOW25", "WELCOME30" and "FREESHIP", none of which had ever existed. The answers cited two coupon aggregator pages. They also offered a real influencer code worth 20% off that had leaked. The store published an official offers page stating that its only public offer was 10% off the first order through its newsletter. It retired the influencer code, issued unique codes to partners and asked both aggregators to remove the fake codes. It also changed its invalid-code message to link to the offers page. Two months later, most answers described the newsletter offer, and code-related tickets had fallen by more than half.

What do people ask about AI and coupon codes?

Should we create the codes assistants invent?

No. That rewards invented answers and trains shoppers to expect them. Publish your real offers clearly instead.

Can we stop coupon sites listing our store?

Usually not, but many will remove fake or expired codes on request, and some let merchants verify listings.

Do invented codes affect our search rankings?

Not directly. They affect conversion and trust, which is why a clear offers page and a helpful checkout message matter.