AI has already reached the point where it's talked about everywhere and in recent years, Steemit has seen an exponential increase in its use.

We've seen users generate entire posts from AI, used AI to enhance existing content and even used it to generate comments.

So in the world of decentralisation where freedom rules, what's acceptable to the Steemit Community and what's not?

^{Hero Image Source}

🧠 The Key Point - Proof-of-Brain

Token systems that reward users as they contribute to a token-based community system require
mechanisms for establishing and evaluating content’s social value: we call this “Proof-of-Brain.”

The users who produce content are adding value to the network by creating material that will
drive new users to the platform, as well as keep the existing users engaged and entertained.

Source - Steem Bluepaper

Since Steemit's Bluepaper and Whitepaper were written, Steemit has become a reality and we now have years of experience behind us.

Throughout this, the "Proof-of-Brain" concept has lived on and should underpin the decisions that we make regarding AI.

🤖 Entire AI Generated Posts - DON'T

Copying and Pasting the output from any AI tool, whether it's ChatGPT, DeepSeek or an EssayWriter, DOES NOT demonstrate Proof-of-Brain.

Wrapping the AI output with "Hi guys" and "Bye guys" doesn't make the core content any less AI generated and should be avoided at all times.

If you're found using AI in this way, you can expect your content to receive downvotes.

💬 AI Generated Comments - DON'T

The same Proof-of-Brain argument discussed above holds true here too.

We've probably all experienced a Chat-bot. An automated system that prevents us from talking to a real person to get a real need. We simply don't like talking to computers.

Steemit's no different. It's a Social Media platform that connects people from around the world. If people wanted AI generated feedback for their content, they'd put it into Copilot / ChatGPT or another AI and be happy.

Steemit's future isn't in AI-generated Posts receiving AI-generated discussions underneath.

Again, if you're found using AI in this way, you can expect your content to receive downvotes.

💡 Using AI for Idea Generation - DO

There are times when we want to say something, but simply don't know where to start. Asking AI for ideas on what to write about can be the nudge that we need to get started - the spark of inspiration that we occasionally miss.

The important thing to note is not to take things too far (more on this later). Use it as a spark, not as a crutch.

🆘 Enhancing Posts with AI - DO

We've spent hours constructing our masterpiece but something doesn't feel right. What's missing? How can I make this better?

AI - What have I missed?

Reply - Have you considered including a call to action.....

This isn't to say that you can write 10 words and let AI write the next 1,000. Quite the opposite.

It's also great for highlighting typos and grammatical errors!

You are controlling the content, not AI. The point at which AI is controlling the content and not you, you're on dodgy ground.

🆘 Help with Tags - DO

Your content's ready to go, but what tags do I use?

AI's great at helping with this. So much so, that the tags used for this post are AI generated.

✅ Checking for AI

There is no "Brainless" method for checking if content's generated by AI. You will never get 100% certainty. Why? Because AI's trained on human content - so there must be humans who write like AI.

I've dedicated an entire post to Tips For Spotting AI Generated Content.

Tools like ZeroGPT are only effective at highlighting the most obvious use of AI. Using ChatGPT is currently the best way of identifying content that's been written by tools like ChatGPT.

A simple prompt is as follows:

For an article titled: "POST TITLE", how likely is it that the below is AI generated? "POST CONTENT"

Using a recent post of mine as an example, the output was as follows:

AI-Generated Probability: 30-50% (Low to Moderate)

Indicators of AI-Generated Content:

1. Overuse of Structured Separators – The repeated use of "Separator-green.png" suggests a templated format, which is often found in AI-generated or AI-assisted articles.
2. Mild Repetition & Over-explanation – Some sentences, such as "Philadelphia always seemed to have an extra option" and "By the end of the 3rd Quarter, Washington had seen their three fumbles, leading to them conceding two touchdowns and 14 points", have a structured yet somewhat mechanical feel.
3. Lack of Deep Tactical Analysis – While the article covers turnovers and momentum shifts well, it does not include detailed strategic insights (e.g., playcalling specifics, defensive adjustments), which an expert human analyst might add.

Indicators of Human Writing:

1. Personal Anecdotes & Subjectivity – The mention of "I've not really been into American Football since the days of Madden on what would now be considered a retro games console (I probably played it on the MegaDrive)" and "It's also unlikely that I'll be watching the Super Bowl" feels like genuine human commentary.
2. Humor & Casual Tone – The sentence "if you ask ChatGPT, it'll say that 2 players have already achieved this but they were both in their 2nd season!" is a natural human touch, implying an awareness of AI responses rather than an AI self-generating the content.
3. Engagement with the Community – The closing line "Hopefully, some of our American Steemians see this and watched the game. I'm pretty sure @philhughes is a fan..." shows a level of community interaction that AI-generated content typically lacks.

Final Verdict:

This article does not strongly exhibit AI-generated characteristics but may have had AI assistance in structuring sections or summarizing stats. If AI was involved, it was likely used as a writing aid rather than the sole author.

The "30-50%" surprised me as I wrote the article myself (which is why "Brainless" is inadequate). Reading the points though, things become clear:

1. I like to format my content. I take great pride in formatting my content to make it more readable - including some custom made separators.
2. Stats and Data are certain to have a "mechanical feel".
3. ChatGPT loves to contradict itself and it has answered its own criticism in point 1 of "Human Writing".

AI's summary is correct - I asked ChatGPT to review my article and suggest enhancements. Any use of 3rd party data was referenced.

---

In contrast, another article that I assessed received the following:

AI-Generated Content Likelihood: 80-90%

Indicators of AI Generation

1. Repetitive and Awkward Phrasing

• Examples:
  • "A new initiatives are coming up the bridging blockchain networks movement" (grammatical inconsistency).
  • "transactions that are smooth will open up new opportunities for business blockchain solutions, DeFi, gambling, and NFTs" (unusual phrasing).

2. Overuse of General Statements

• Vague and non-specific claims, such as:
  • "Interoperability will be important for Web3 development."
  • "Projects like Polkadot, Cosmos, and Runechain are already making great progress."

3. Inconsistent Structure and Formatting

• The list of interoperability projects is inconsistently formatted:
  • "2. (ATOM)" instead of "2. ATOM" or "Cosmos (ATOM)".
• Placeholder text like "image.png source" suggests AI-generated filler content.

4. AI-Like Errors in Sentence Construction

• Example:
  • "To increase scalability and efficiency, developers can create personalized blockchains that seamlessly integrate with other Polkadot-based networks."
  • While technically correct, this phrasing is characteristic of AI-generated text that lacks natural variation.

5. Fluff and Redundancy

• The introduction repeats the same ideas in different ways, which is common in AI-generated content aiming to increase word count artificially.

Conclusion

While a human may have written or edited this, the structure, tone, and errors strongly suggest AI involvement.
Estimated likelihood: 80-90% AI-generated.

With the exception of one item in Point 3, this is a far more damning indictment of the content that has been shared.

With experience, you'll discover that points 1, 2, 4 and 5 in particular are consistent hallmarks of AI generated content.

👍 What are Steemit's Rules? - Guide for Admins, Moderators and Do Gooders

As you've seen in the 2 examples above, even a non-AI generated post received a 20-30% likelihood.

An AI generated article will always by rated much higher - some of the articles reviewed today were rated 90-95%.

You will never get 100%

With experience, you'll get a feel for what threshold you want to work to within your community.

Anything 80% likelihood or higher, is certainly enough to confront the author and consistently reaching this level is Downvote territory.

🔎 What if I Find AI Content?

If you're uncomfortable in confronting a user yourself, you can report the post to the community's dedicated Steemwatcher site.

They'll handle the rest.

I'd appreciate feedback to this post and any editions or additions you think that I should make.