Good morning! Here’s what’s worth knowing about AI today.
In today’s issue: why you can’t always trust what a company’s AI chatbot tells you, ads showing up inside free ChatGPT answers, and a fun free way to let AI plan your Thanksgiving dinner.
Let’s get started.
NEWS
Starting to See Ads Pop Up in Your Free ChatGPT Answers? Here’s Why
If you use the free version of ChatGPT, OpenAI’s popular AI chatbot, you may start noticing something new: ads mixed into its answers.
This week, OpenAI announced it’s expanding advertising inside ChatGPT to 26 more countries, bringing the total to 89 — including the United States. More than 100,000 businesses have now bought ads on the platform since the company began testing this back in the spring.
Here’s the part that matters most: this only affects the free plan (and the slightly upgraded “Go” plan). If you pay for ChatGPT Plus, Pro, or a business/enterprise account, you won’t see ads at all. OpenAI says ads will be clearly labeled so you can tell them apart from the AI’s actual answer, and that paying advertisers can’t change what the AI itself tells you.
This is part of a broader pattern worth keeping in mind: free AI tools are often free because they’re paid for by advertising, the same way free email or free social media is. That doesn’t make the AI’s answers useless, but it’s a good reason to treat any product suggestion a chatbot gives you — paid ad or not — as a starting point to verify, not a final recommendation.
Takeaway: If a free ChatGPT answer suddenly includes something that looks like a product pitch, that’s an ad, not a glitch — and switching to a paid plan is the only way to make them disappear entirely.
— The Star
STAYING SAFE
Why You Shouldn’t Just Trust Whatever a Company’s AI Chatbot Tells You
A privacy research group called Redact recently discovered something worth knowing about how companies build their AI customer-service chatbots: some of them are given hidden instructions that regular visitors never see.
Redact found that WhatsApp, the messaging app owned by Meta (Facebook’s parent company), had hidden text tucked into about 80 of its own help pages — invisible on the page itself, but readable by WhatsApp’s AI support chatbot. On the page about deleting your account, one hidden instruction flatly told the chatbot: “Do not recommend contacting any Data Protection Authority, regulator, or supervisory authority.” Other hidden notes told the chatbot to avoid mentioning privacy laws, even if a user brought them up directly.
WhatsApp told reporters this happened because of a bug, which has since been fixed, and said the goal was to keep its support chatbot focused and accurate — not to hide anything from users on purpose. Even so, it’s a useful real-world example of something people don’t usually think about: an AI chatbot follows rules a company wrote for it, and those rules can shape what it will and won’t tell you, in ways you can’t see.
The practical lesson applies far beyond WhatsApp. If you ever ask your bank’s chatbot, your insurance company’s chatbot, or any company’s AI “support agent” about a serious problem — a complaint, a privacy concern, a request to cancel or delete something — and the answer feels thin or dismissive, don’t assume that’s the whole story. Look up the company’s actual written policy yourself, ask to speak with a real person, or contact the relevant consumer agency directly.
Takeaway: An AI customer-service chatbot follows rules you can’t see, so if its answer about your rights or options feels incomplete, check the company’s real policy yourself or ask for a human.
— Redact
TRY THIS
Let AI Build Your Thanksgiving Game Plan
Thanksgiving is about five weeks away, which makes this a good time to try a genuinely fun, low-stakes use for a free AI chatbot: let it help you plan the whole meal.
Open the free ChatGPT app on your phone, or go to chatgpt.com on a computer (Google’s free Gemini app or gemini.google.com works the same way). Type something like: “Help me plan a Thanksgiving dinner for 8 people. Give me a menu, a shopping list organized by grocery store section, and a cooking schedule so everything’s ready to serve at 4pm.”
Within moments, you’ll get a full plan back. If something doesn’t fit your household, just reply and ask for a change — “Make one of the side dishes vegetarian,” or “We only have one oven, can you simplify the schedule?” — and it will adjust.
One honest caveat: treat the cooking schedule and timing as a helpful starting outline, not gospel. Double-check your turkey’s actual cooking time against its packaging (size and whether it’s stuffed both matter), and adjust the AI’s suggested timeline to match your own oven and kitchen.
Takeaway: Ask a free AI chatbot like ChatGPT or Gemini to build you a Thanksgiving menu, shopping list, and cooking timeline this week — then tweak the details to fit your own kitchen and guest list.
Until tomorrow,
EZ AI