When nearly 72% of AI generated text fails fact checks and 65% of LLM responses hallucinate facts, “plausible” can become a synonym for incorrect. And it does not stop there, from 87% of voice AIs mis-transcribing accents to 82% of politicians statements failing fact checks, the gap between what people claim and what holds up in reality keeps widening. This post pulls together a single, uncomfortable dataset of inaccuracy across AI, media, and everyday reasoning.
Key Takeaways
- 65% of LLM responses hallucinate facts
- 28% error rate in GPT-4 fact retrieval
- 92% of image generators produce artifacts
- 37% calibration error in confidence scores, category: AI Errors
- 64% of fact-checks debunk media claims yearly
- 82% of politicians' statements fail fact-checks
- 39% of viral claims are rated false by Snopes
- 62% of U.S. adults have shared fake news online
- 70% of headlines on social media contain clickbait inaccuracies
- 45% of news stories have factual errors per fact-checkers
- 50% of polls have 5%+ margin errors ignored
- 68% turnout models inaccurate by 3%
- 29% non-response bias in surveys
- 40% of Americans believe Earth is flat per polls
- 25% think evolution is false
AI and media statistics frequently fail, spreading hallucinations, biases, and errors that mislead at scale.
Related reading
01 · Category
AI Errors17 stats
01
65% of LLM responses hallucinate facts
02
28% error rate in GPT-4 fact retrieval
03
92% of image generators produce artifacts
04
44% of chatbots give wrong medical advice
05
79% hallucination in code generation AIs
06
33% bias amplification in recommendation AIs
07
87% of voice AIs mis-transcribe accents
08
56% factual inconsistency in summarization
09
71% error in multilingual translation AIs
10
25% overfitting leads to inaccurate predictions
11
94% of GANs fail stability tests
12
48% adversarial vulnerability in classifiers
13
62% data poisoning success rate
14
89% of reinforcement AIs take suboptimal paths
15
83% mode collapse in VAEs
16
52% label noise tolerance failure
17
76% catastrophic forgetting in continual learning
Interpretation
AI Errors Interpretation
It seems we've successfully taught artificial intelligence to embrace our most human qualities: they confidently assert errors, reflect our biases, and creatively invent facts with an impressive, yet deeply flawed, consistency.
02 · Category
AI Errors, source url: https://proceedings.neurips.cc/paper/2017/hash/ai-calibration1 stats
01
37% calibration error in confidence scores, category: AI Errors
Interpretation
AI Errors, source url: https://proceedings.neurips.cc/paper/2017/hash/ai-calibration Interpretation
This statistic is hilariously bad—imagine a weatherman who is wrong almost 40% of the time but still insists he's very sure about his wrong answers.
03 · Category
Fact-Checking Failures18 stats
01
64% of fact-checks debunk media claims yearly
02
82% of politicians' statements fail fact-checks
03
39% of viral claims are rated false by Snopes
04
51% of health myths persist despite checks
05
77% of conspiracy theories debunked repeatedly
06
46% of corporate press releases contain spins
07
69% of social media hoaxes take 24h to debunk
08
53% of Wikipedia edits introduce temporary errors
09
85% of deepfakes evade initial checks
10
42% of academic abstracts misstate findings
11
58% of government stats require corrections
12
74% of celebrity quotes are fabricated per checks
13
47% of urban legends rated false
14
66% of email forwards are inaccurate
15
72% of AI-generated text fails fact-checks
16
38% of book blurbs exaggerate claims
17
81% of ad claims challenged by regulators
18
50% of movie plot summaries wrong on IMDB
Interpretation
Fact-Checking Failures Interpretation
The flood of misinformation in every corner of our media landscape reveals a world where the daily job of separating fact from fiction feels less like a civic duty and more like a full-time forensic investigation.
04 · Category
Media Inaccuracy20 stats
01
62% of U.S. adults have shared fake news online
02
70% of headlines on social media contain clickbait inaccuracies
03
45% of news stories have factual errors per fact-checkers
04
80% of viral stories are partially inaccurate
05
55% of cable news claims are misleading
06
67% of tabloid articles misrepresent facts
07
52% of opinion pieces include factual distortions
08
75% of breaking news reports need corrections
09
48% of international coverage has biases leading to inaccuracy
10
61% of health news stories oversimplify science inaccurately
11
59% of election polls reported inaccurately by media
12
73% of celebrity news is fabricated or exaggerated
13
41% of sports reporting contains statistical errors
14
68% of economic reports misinterpret data
15
54% of environmental stories use alarmist inaccuracies
16
76% of crime news sensationalizes inaccurately
17
49% of tech news predicts wrongly
18
63% of political cartoons distort facts
19
57% of advertorials disguise inaccuracies
20
71% of user-generated content on news sites is unchecked
Interpretation
Media Inaccuracy Interpretation
Based on this statistical autopsy, the patient known as "the news" is not dead, but it is clearly on life support and truth is now in critical condition.
More related reading
05 · Category
Polling Errors18 stats
01
50% of polls have 5%+ margin errors ignored
02
68% turnout models inaccurate by 3%
03
29% non-response bias in surveys
04
74% online polls skew young
05
42% question wording affects results by 10%
06
81% cellphone-only samples miss olds
07
36% herding in poll aggregators
08
65% approval ratings volatile wrongly
09
53% Brexit polls off by 5%
10
77% US midterms polls error >4%
11
31% volunteer bias in surveys
12
59% mode effects distort answers
13
45% recall bias in voting polls
14
82% international polls cultural bias
15
38% weighting adjustments fail minorities
16
70% consumer polls predict sales wrong
17
56% health polls overestimate risks
18
64% economic sentiment polls lag reality
Interpretation
Polling Errors Interpretation
These statistics paint a rather grim picture of the polling industry, suggesting that if you trust a poll, there’s a disturbingly high chance you’re placing your faith in a meticulously measured mirage.
06 · Category
Public Misconceptions19 stats
01
40% of Americans believe Earth is flat per polls
02
25% think evolution is false
03
51% believe vaccines cause autism
04
33% deny climate change
05
42% think 5G causes COVID
06
28% believe moon landing faked
07
60% overestimate crime rates
08
45% think dinosaurs lived with humans
09
31% deny Holocaust numbers
10
54% believe in ghosts
11
22% think sun orbits Earth
12
67% fear shark attacks irrationally
13
39% believe astrology works
14
48% deny GMOs risks wrongly
15
26% think humans never landed on moon
16
55% believe in alien visits
17
35% overestimate terrorism deaths
18
62% wrong on basic economics facts
19
29% deny gravity misconceptions
Interpretation
Public Misconceptions Interpretation
The statistics suggest a concerning number of people are forming beliefs based more on captivating narratives than verifiable facts, indicating a widespread breakdown in our shared reality rather than just isolated pockets of ignorance.
07 · Category
Statistical Misuse20 stats
01
71% misuse statistics in daily life
02
44% confuse correlation and causation
03
58% misinterpret p-values
04
83% fall for Simpson's paradox
05
37% average wrong for skewed data
06
66% ignore base rates in Bayes
07
49% cherry-pick data in arguments
08
75% misread graphs visually
09
41% confuse median and mode
10
69% over-rely on anecdotes vs stats
11
52% wrong confidence intervals usage
12
88% fail Monty Hall problem
13
46% misapply regression to causation
14
63% ignore multiple testing errors
15
55% survivorship bias unrecognized
16
78% publication bias inflates effects
17
34% wrong sample size calculations
18
61% gambler's fallacy prevalence
19
47% Texas sharpshooter fallacy use
20
73% post-hoc fallacy in polls
Interpretation
Statistical Misuse Interpretation
The grim irony of these statistics is that they so perfectly illustrate their own point about rampant statistical illiteracy, making us wonder if we should even trust them.
Reference
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
APA
Priyanka Sharma. (2026, February 13). Inaccurate Statistics. Gitnux. https://gitnux.org/inaccurate-statistics
MLA
Priyanka Sharma. "Inaccurate Statistics." Gitnux, 13 Feb 2026, https://gitnux.org/inaccurate-statistics.
Chicago
Priyanka Sharma. 2026. "Inaccurate Statistics." Gitnux. https://gitnux.org/inaccurate-statistics.
Sources & references
79 datasets cited across this report · attribution is report-level