Frequently asked questions

The direction label describes the dominant pattern of signals found in that specific article — tone, framing, source selection, and omission — not a political judgment about the outlet, author, or subject matter. The same publication can receive different labels across different articles depending on how a story is assembled.

The confidence score reflects how clearly and consistently the directional signal appeared. A score of 0.85 means the signals were clear and repeated across the article. A score of 0.30 means signals were present but faint or mixed. Read it alongside the direction label — 0.70 on a Center result means something different from 0.70 on a Left-leaning result.

Yes. Bias direction describes framing patterns, not factual accuracy. A well-reported article can still use language, sourcing, or emphasis that leans in one direction. NeutralEye is not a fact-checker — it analyzes how a story is constructed, not whether its claims are true.

Because NeutralEye analyzes the text you submit, not the outlet's overall reputation. A wire report and an opinion column from the same publication can produce very different results. That's intentional — labeling by outlet would be a shortcut that ignores how individual stories are actually written.

Confidence measures how clearly and consistently the bias signals appeared across the submitted text. A high confidence score means tone, framing, sourcing, and attribution all pointed in the same direction throughout the article. A low score means signals were mixed, sparse, or contradictory.

No. A well-written opinion column can score high confidence because its rhetorical structure is deliberately consistent — that's not a flaw. High confidence means the pattern was clear, not that the article is manipulative. What matters is reading the direction, the score, and the evidence together.

Not exactly. Low confidence can mean the article is genuinely balanced and signals cancel out. But it can also mean the text was too short, the writing was inconsistent, or the story was still developing when filed. A low score is a flag for caution, not a clean bill of health.

Think of confidence as a volume dial, not a pass/fail gate. A high-confidence Left-leaning result with multiple quoted examples carries more weight than a high-confidence label with thin evidence. The direction tells you which way; the confidence tells you how strongly; the examples tell you why.

Article text submitted for analysis is processed to generate a result and is not permanently stored by default. Signed-in users can opt in to saving their analysis history — this stores the result and metadata, not the full article text.

For anonymous users: only the data needed to enforce rate limits (IP address, request count). For signed-in users: your email, analysis history if you opt in, and daily usage count. We do not sell data or share it with advertising networks. See the Privacy Policy for the full picture.

No. The extension only activates when you click the NeutralEye icon. It reads the current tab's content only at that moment — it does not run in the background, track your browsing, or access tabs you haven't explicitly submitted for analysis.

No. Submitted text is used only to generate your analysis result. It is not retained for model training or used to improve the underlying AI beyond the current session.

Free users get 5 analyses per minute and 10 per day through the Chrome extension. The web analyzer has a separate rate limit of 5 requests per minute. No account is required to get started — limits apply per IP address for anonymous users.

Pro gives you higher daily limits, saved analysis history, and access to Compare Analyses — which lets you run two articles side by side and see where framing diverges. Pro is billed monthly. Stripe integration is coming soon — join the waitlist on the pricing page.

The extension and website have separate rate limits. Extension analyses count against your extension daily quota; web analyses count separately. Signing in links both to your account so history saves across both.

No. You can run analyses on the website without an account, subject to the anonymous rate limit. An account is needed to save history, use the extension with authenticated limits, and access Pro features.

It checks five signal families together in one pass: tone (emotional loading, verb choices), framing (what the article centers vs. backgrounds), attribution (who gets quoted and how), source balance (distribution and credibility of sources), and omission (context that's missing). The direction emerges from the pattern across all five — not any single signal in isolation.

It works best on articles: news reports, analysis pieces, opinion columns, and editorials. Very short texts, paywalled pages, navigation-heavy pages, and non-article content (product pages, forums) may produce unreliable results. The system flags these cases when it detects them.

NeutralEye surfaces patterns — it doesn't claim certainty. Satire, irony, and unusual writing styles can resemble bias signals. The analysis should be treated as a structured second opinion that prompts closer reading, not as a definitive verdict. The evidence section exists precisely so you can check the reasoning yourself.

For URL submissions, the system attempts to extract article text. Paywalled pages that block extraction will return an error. You can always paste the article text directly if you have access — that bypasses extraction entirely.