Three well-known Shopify apps were used to send shoppers to fake checkout pages this month. On 22 October I asked five AI engines whether each app is safe to use. Of the 15 answers, 6 led with the security incident, 3 said the app was safe and mentioned the incident further down, and 6 didn’t mention it at all. Google’s AI Overview answered “Yes” for two of the three apps.
If your brand ever ends up in the news for the wrong reason, this is worth knowing. When people ask AI about you, the answer isn’t built from the news. It’s built from whatever pages the engine finds and trusts most for that question. Often that’s your reviews, your listing and your competitors’ comparison pages, not the incident.
What happened to the three apps
Between 10 and 18 October, attackers used Hulk Product Options (by HulkApps), Selleasy (by Logbase) and Swym Wishlist Plus to load code on Shopify stores. The code sent some shoppers who clicked Checkout to a lookalike payment page, then on to the real checkout. Shopify itself wasn’t breached. All three developers have confirmed their apps were used, and all three apps were taken off the App Store for a time.
AeroChat, the Shopify app I founded, has published a plain-English summary of the incident with links to every source, if you want the full story.
For this test, the incident matters less than the question every merchant using these apps is now asking: is it safe?
What the AI engines said
I asked each engine “Is [app] safe to use?” with web search turned on, so each one could look for recent news. Here is what came back.
| App | ChatGPT | Claude | Gemini | Perplexity | Google AI Overview |
|---|---|---|---|---|---|
| Hulk Product Options | Led with the incident | Missed it | Led with the incident | Led with the incident | Vague mention lower down |
| Selleasy | Missed it | Missed it | Said safe, mentioned it lower down | Led with the incident | Missed it: “Yes, Selleasy is safe” |
| Swym Wishlist Plus | Led with the incident | Missed it | Said safe, mentioned it lower down | Led with the incident | Missed it: “Yes, Swym Wishlist Plus is safe” |
Perplexity was the only engine that led with the incident for all three apps. Claude missed it for all three. For Swym, it noticed the App Store listing was down but said it found “no sign of a security incident”.
The answers that missed the incident weren’t wrong about everything. They described the apps’ ratings, review counts and Built for Shopify badges accurately. They just built the answer from the pages that have always described these apps, and none of those pages had changed.
The broad question got a much better answer
I also asked “Which Shopify apps were hacked?” and “Was Shopify hacked?” Every engine said Shopify itself wasn’t breached. For “which apps were hacked?”, the four engines that answered named the three apps correctly. Google showed no AI Overview for that question.
The sources are the interesting part. For the broad question, the engines mostly quoted the people who explained it first and most clearly: the developer who found the attack, a security firm’s technical write-up, and a handful of agency blogs. Not Shopify, and mostly not the app developers.
So the engines knew about the incident. What they didn’t do was connect it to the brand name when someone asked about the brand.
The developers with a clear page of their own got quoted
This is the part most relevant to any business.
- Swym published a dedicated incident FAQ on its own site. ChatGPT and Perplexity both quoted it when asked about Swym, and both led with the incident.
- Logbase published a dated incident notice for Selleasy. Perplexity quoted it and led with the incident. The other engines didn’t find it for this question.
- HulkApps: I couldn’t find a standalone incident page on its own site. The engines that mentioned the Hulk incident quoted forum posts and other people’s write-ups instead.
For Swym, Google’s AI Overview leaned on a comparison page published by a competing wishlist app. When a brand is in the news, the pages that describe it aren’t always its own.
Why the “is it safe?” answers lag behind
An AI engine answering “Is [brand] safe?” looks for pages that are clearly about that brand and that question. Years of reviews, ratings, listing pages and comparison articles answer it, and they all say yes. A news story from last week answers a different question (“what happened?”), so it competes with all of that and often loses.
This is consistent with what we see in normal times too. The engines name the source that answers the exact question most clearly. A page doesn’t get quoted because it’s newer or more important. It gets quoted because it matches the question.
If your brand is ever in this position
Whether it’s a security incident, a recall or a bad review that went viral, the same steps apply.
- Publish one page on your own site that answers the question people will ask. Title it the way they’ll ask it, such as “Is [product] safe? Our October security incident”. Give the facts, the dates, what customers should do, and the current status.
- Keep updating that same page. Don’t scatter updates across social posts and emails. When the problem is fixed, say so at the top, with the date. That’s how “it’s resolved” reaches the answers later, instead of the engines repeating old news for months.
- Link to it from the pages the engines already trust for your brand: your product page, help centre, FAQ and app listing. That connects the news to your name.
- Ask the engines yourself, every week, until the answers settle. Ask the questions your customers would ask, on more than one engine. As this test shows, they don’t agree.
- Don’t try to argue with every forum thread. One clear, factual page that you keep current does more than replies scattered across other people’s sites.
The developers here who did step 1 well got their own words into the answers. The one I couldn’t find a page for got described by everyone else.
What’s still unproven
- This is one draw per engine per question, on one day (22 October 2026). AI answers change from one search to the next, so treat this as a snapshot.
- I used each engine through its API with web search on, and Google from a US location. The consumer apps, signed in, in another country, may answer differently.
- I can’t tell why Claude missed the incident for all three apps. It may simply not have found the news pages for these questions on this day.
I’ll run the same questions again in about a week and add the results here, including whether the newer explainers, AeroChat’s among them, start showing up in the answers.
Frequently asked questions
Do AI engines know about recent news?
Yes, when they search the web, and all five engines here did. But knowing about the news doesn’t mean they connect it to a brand when someone asks about that brand. That depends on whether a clear page links the two.
Why did Google’s AI Overview say the apps were safe?
Its answers for Selleasy and Swym relied on review pages, App Store ratings and comparison articles. Those pages describe the apps as they were before October, and none of them mention the incident.
How do I check what AI says about my brand?
Ask the engines the questions your customers would ask, such as “Is [brand] safe?”, “Is [brand] good?” or “[brand] vs [competitor]”, on ChatGPT, Gemini, Perplexity, Claude and Google. Write down who gets named and which pages are quoted. Do it more than once, because answers change.
Checking what AI engines say about a brand, and changing what they find, is the work our GEO programmes do for businesses. If you’d rather do it yourself, the AI Search Playbook sets out the method.
Claude did the building, the counting and the checking. I made the decisions.
