By Cindy Gordon, selective visibility strategist and business mentor
You may have noticed content on Instagram and TikTok getting flagged with a Made with AI label over the last few weeks. I have not been flagged personally, but I am seeing more of it in my feed, and sometimes on content that is very clearly not AI generated.
Something specific happened on August 2nd that explains it, and it is going to keep affecting how your content gets received. Here is what actually changed, and the part that matters more than the label itself.
What changed on August 2nd?
A set of rules called Article 50 of the EU AI Act took effect.
The short version: AI companies now have to make it possible to identify when content was generated by AI. Not label it visibly, necessarily. Just make it identifiable.
Does the EU AI Act apply to businesses outside Europe?
Functionally, yes.
It applies to any AI system whose outputs might reach someone in Europe. Which, if you are publishing on the internet, means you. The tools you use are complying globally rather than building separate versions for different continents.
Does Claude put watermarks on AI generated text?
Yes. Anthropic, the company behind Claude, began embedding invisible watermarks in Claude generated text on August 2nd, across all of its products.
You cannot see them. I cannot see them. They do not change the meaning or readability of the writing. But they survive copying and pasting, and platforms with the right tools can detect them. Heavy editing, translation, or converting between formats can strip them out.
Other AI companies are building similar systems. This is not a one company decision, it is the direction the whole industry is moving.
Why is content that is not AI getting flagged?
Because the detection systems are new, and new systems are not accurate systems.
TikTok and Instagram both use a fairly basic technology that tries to detect and label AI content automatically. It is already producing false positives. I am seeing reports of TikTok flagging content as AI generated that had no AI involved at all.
Sometimes it happens because a creator used a design tool for one small edit, the platform picked up that tool’s digital fingerprint in the file, and the whole video got marked. Creators are publicly frustrated about content being labeled AI when they wrote every word of it themselves.
Worth knowing: the unreliability runs both directions. Anthropic has said plainly that finding a watermark does not prove Claude wrote something, because people use it to edit and translate their own work. And not finding one does not prove a human wrote it, because older models were not watermarked and short passages may not carry a detectable mark.
The label is not a reliable signal right now. It is an overaggressive early version of a policy that is going to keep evolving.
Does the Made with AI label actually hurt your business?
The badge itself is not the thing that makes or breaks you.
Whether your reel picks up a small Made with AI tag is not what decides whether someone hires you. If that is the part you are worried about, you are worried about the wrong part.
Will AI search engines stop recommending AI written content?
This is the question I would actually focus on.
Right now, when someone asks ChatGPT or Perplexity who they should hire, or which business is best for something, the AI pulls from a lot of sources. If AI generated content starts getting flagged in that process and weighted as less trustworthy, that has real business consequences.
Nobody knows exactly how this shakes out. But the direction is getting clear: AI systems increasingly want to recommend content a human actually wrote, because it is verifiable in a way AI generated content is not.
The thing Anthropic said about its own AI
This part made me laugh. Anthropic acknowledged on its own website that AI models are, and I am quoting directly, “fond of the construction, ‘This isn’t X, it’s Y,'” and that they use the word “quietly” far more than you would expect.
The company that makes Claude is telling you exactly what to look for in AI generated writing. I love it. And it confirms what a lot of us have been noticing for a while about how this stuff sounds.
Is it okay to use AI for your business content?
Yes, and here is the line.
The real question is not whether your content gets labeled. It is whether you are using AI to draft, or letting it write for you.
Using AI for research, for outlines, for brainstorming, for organizing your notes: that is smart. That is what it is genuinely good at, and I use it every single day that way.
Letting AI write the words that go out under your name is a different thing entirely. Not because of the label. Because it is not you, and people can tell. And increasingly, so can the systems deciding whether to recommend you.
Being the answer AI can trust means being the human answer, not the AI flavored version of one.
Why this is really a findability problem
Underneath all of it, this is a findability question. Are you the answer that AI systems and real people can confidently point to?
Your assignment this week
Open the last three things you published. A caption, an email, a blog post, whatever they were.
Read them out loud and ask one honest question: did I write this, or did I let AI write the whole thing in my voice?
If it is the second one, that is your work. Not because a label is coming for you, but because being the answer people and AI trust means being the actual person behind the words.
The label is not the risk. The risk is sounding like everyone else.
Prefer to listen?
Which door are people
finding you through?
Someone decides to hand you money one of three ways. She sees you and remembers. The algorithm shows you to her. Or an AI tool says your name when she asks who to hire.
Most businesses are strong in one of those and blind in the other two. That is usually why the effort and the results do not match.
The Findability Audit is free. Twelve questions, and you will know which layer is costing you clients, and which one is quietly making you money without you knowing why.
Frequently Asked Questions
Platforms including TikTok and Instagram began automatically detecting and labeling AI content after Article 50 of the EU AI Act took effect on August 2, 2026. The detection systems are new and produce false positives. Content can be flagged because a design tool left a digital fingerprint in the file, even when no AI was used to create it.
Article 50 is the transparency section of the EU AI Act, effective August 2, 2026. It requires AI companies to make it possible to identify when content was generated by AI. Although it is European legislation, it applies to any AI system whose outputs could reach someone in Europe, so most major AI tools are complying globally.
Yes. Anthropic began embedding invisible watermarks in Claude generated text on August 2, 2026, across its products. The marks are not visible to readers and do not change the writing, but they survive copying and pasting and can be detected by platforms with the right tools. Heavy editing, translation, or format conversion can remove them.
The label itself is not what determines whether a business succeeds. The more significant risk is what happens as AI recommendation engines begin weighting AI detected content as less trustworthy when deciding which businesses and people to recommend to someone asking for a referral.
Using AI for research, outlines, brainstorming, and organizing notes is a strong use of the tool. The line is letting AI write the finished words that publish under your name. That reads as generic to people, and increasingly to the AI systems deciding who to recommend.
Read it out loud. Anthropic has publicly noted that AI models favor the construction “This isn’t X, it’s Y” and overuse the word “quietly.” Beyond specific tells, the question is whether the piece sounds like the person whose name is on it or like a competent generic version of that person.
About Cindy Gordon
Cindy Gordon is a selective visibility strategist and business mentor for female entrepreneurs. A 6x founder who has built six businesses and sold four, she holds a Masters in Special Education with a focus in Behavior Analysis and brings the discipline of individualized assessment to visibility work. She has grown a brand’s following to a million, built a 900,000 person community and a 250,000 person email list, and led a team of 11. As a business coach, she helps women diagnose what is breaking their content, where their visibility actually belongs, and how to show up with clarity instead of noise. Cindy is the founder of Exclusively Cindy and the host of The Strategic Entrepreneur podcast, where she explores selective visibility, proof of human, and what it takes to become unmistakable.
