Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallNo—Sam Altman did not claim that social media is demonstrably mostly bots. In a September 8, 2025 post discussed by TechRadar, he described AI-heavy Twitter and Reddit as feeling “very fake” and said he now assumes what he reads may be fake or bot-generated. He offered possible explanations, not a measured bot percentage.
Independent OpenAI case studies document real, AI-assisted influence operations on X, Facebook and Bluesky. They show that generated personas and coordinated posting exist, but their small, low-engagement samples cannot be used to estimate how much of social media is automated.
As an Amazon Associate I earn from qualifying purchases.
What Sam Altman actually said
His September 2025 observation
Altman wrote that “AI Twitter/AI Reddit feels very fake in a way it really didn’t a year or two ago.” He added: “I have the strangest experience reading this: I assume it’s all fake/bots.”
Those are personal impressions. Altman did not say that a specific share of posts or accounts are bots, and he did not present a platform-wide measurement.
#1 Best Overall
The mechanisms he suspects
Altman pointed to several overlapping causes:
- Real people may have adopted recognizable LLM-style phrasing.
- The extremely online population often reacts in correlated ways, making independent opinions look coordinated.
- Online discussion swings through repeated “it’s so over” and “we’re so back” hype cycles.
- Platforms optimize for engagement, while creator monetization rewards attention, encouraging sensational or repetitive posting.
These mechanisms can make a human-written feed look synthetic, even when no bot is involved.
Altman’s earlier position on social-media risk
In a June 21, 2023 TIME interview, Altman said social media was “in such a volatile place right now,” said he was nervous about it, and acknowledged that AI’s effects were difficult to predict. He argued that society should discuss AI risks frankly and remain vigilant.
For safeguards, he described the need for “a combination of technical and social solutions to operate in a different way.” That framing is broader than bot detection: it includes platform rules, technical controls and changes in user behavior.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsRank #2
What documented AI influence operations show
Operation A2Z — OpenAI, October 1, 2024
OpenAI described a network operating across X and Facebook. Its operators used models to manage fake personas, write biographies, analyze posts and comments, draft multilingual replies and proofread material before publication.
- OpenAI examined around 150 sampled accounts across the two platforms.
- Typical sampled posts received 0–5 engagements.
- The largest sampled X account had 222 followers; typical followings were in the mid-teens to low twenties.
- OpenAI said the operation showed how models could let operators manage many accounts, but the low engagement limited its measured impact.
“Around 150 accounts” describes this sampled operation only. It is not an estimate of all bots on X, Facebook or the wider internet.
Operation Uncle Spam — OpenAI, June 1, 2025
OpenAI reported a China-origin network that generated polarized U.S. political comments on X and Bluesky, including arguments on both sides of tariff debates.
- Operators asked models for optimal posting times.
- Models generated logos and profile images for fictional personas.
- The operators sought code and methods to extract profile and follower data.
- OpenAI assessed the activity as Category 2 on its Breakout Scale: multi-platform activity with little or minimal breakout.
- Most posts received few or no likes or reposts, and follower counts did not produce matching engagement.
Like A2Z, Uncle Spam demonstrates that AI-assisted coordination is real while also showing that a network can remain small and ineffective.
What this evidence does—and does not—prove
| Question | What the evidence supports | What it cannot establish |
|---|---|---|
| Are AI-operated accounts real? | Yes. OpenAI documented model-assisted personas, content generation and coordination. | That most accounts, or most posts, are automated. |
| Can one operation run many accounts? | Yes. A2Z showed models being used for persona management and multilingual replies. | The scale of every network on a platform. |
| Do these accounts always influence audiences? | Not necessarily. The reported samples generally had very low engagement. | That low-engagement operations are harmless in every context, or that high-engagement posts are authentic. |
| Does AI-like language identify a bot? | No. Altman specifically noted that real people may pick up LLM-like quirks. | A reliable bot decision based on writing style alone. |
| Is there a universal bot percentage? | The cited material does not provide one. | Any platform-wide statistic inferred from these case studies. |
Generated content and automated accounts are also different categories. A human may use an AI tool to draft a post, an operator may control many fictional accounts, and a genuine user may independently write in a formulaic style. Treating all three as the same would overstate what the evidence shows.
Why a social feed can feel fake without being mostly bots
Language convergence
When many people use the same writing assistants, stock phrases, rhetorical structures and emotional cues spread among real users. Similar wording is therefore a clue to investigate, not proof of automation.
Rank #4
- Used Book in Good Condition
Correlated attention
Online communities see the same headlines, recommendation feeds and influencer prompts. Large numbers of people can repeat a frame or slogan because they encountered the same stimulus, not because a single operator directed them.
Engagement incentives
Ranking systems favor reactions, replies and watch time. Monetization adds another reason to post frequently and dramatically. Those incentives can produce exaggerated consensus, outrage or optimism even when every account is human-run.
Recommended Free Tools
Coordinated influence
AI makes it cheaper to create personas, translate messages, generate profile assets and schedule activity across services. The OpenAI reports show this capability in practice, but also show that coordination does not guarantee reach.
Best Value
How to assess a suspicious account or post
- Check the account history. Look for an abrupt creation date, sudden topic changes, copied bios, recycled images or activity across unrelated causes.
- Separate style from behavior. LLM-like wording is weak evidence by itself. Repeated posting patterns, synchronized timing and identical claims across accounts are more informative.
- Inspect engagement quality. A large follower count with few meaningful replies, or many near-identical reactions, deserves caution. Low engagement can also mean the operation had little reach, as in the documented samples.
- Compare independent sources. Find the original document, image or announcement instead of relying on a chain of reposts.
- Check for coordinated amplification. Search distinctive phrases and timestamps across platforms. Cross-platform repetition can indicate organization, but it still requires corroboration.
- Avoid overconfident labels. “Possibly coordinated” is more defensible than calling an account a bot without evidence.
What responses are being considered
Platform and model-provider enforcement
Platforms can remove coordinated inauthentic behavior, label synthetic media, limit bulk automation and publish enforcement reports. Model providers can ban accounts and disrupt operations, as the OpenAI case studies illustrate. These approaches can reduce abuse but depend on consistent detection, appeals and transparency.
User-level verification
TechRadar mentioned Tools for Humanity’s Orb Mini as a possible human-verification response and reported a plan to ship 7,500 devices in the United States by the end of 2025. That report does not establish Amazon availability, a general consumer purchase path or an affiliate program. Hardware-based proof of personhood remains an exploratory option, not a settled solution.
Verification also carries costs: identity and profile data may be collected, access can become unequal for people without the required device, and proving that a person exists does not prove that a post is truthful or independent.
Governance and social norms
Altman’s “technical and social” formulation points to a combined approach: clearer platform rules, independent scrutiny, media literacy, limits on abusive automation and user skepticism about viral claims. No single measure can distinguish every human, assisted and coordinated account without trade-offs.
The defensible conclusion
Altman’s statement is best read as a warning about declining confidence in online authenticity, not as a statistical finding. OpenAI’s A2Z and Uncle Spam reports confirm that AI-assisted influence networks operate across major platforms, while their limited reach and low engagement show why isolated examples cannot be generalized. The useful question is not “What percentage of social media is bots?” but “What evidence shows that this account, message or network is authentic, coordinated or machine-assisted?”
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
