Checker – Full Terms of Service

3.4. Service limits.

• Trial users receive 15 credits; paid plans include the monthly or annual credits specified at the point of checkout.
• When credits are exhausted, you must wait until the next billing cycle or upgrade your plan. We impose no additional rate limits beyond credit exhaustion.
• We do not guarantee continuous, uninterrupted, or error-free operation of the Service. We may suspend or restrict access to the Service for maintenance, security, abuse prevention, legal reasons, or other operational requirements, with or without notice.

5.4. Refunds.

1. Statutory Cooling-Off Period (UK Consumers): Under UK consumer regulations, you typically have a 14-day "cooling-off" period for online purchases. However, by using the Service and consuming your first credit, you consent to the immediate performance of the contract and the immediate delivery of digital content, and you acknowledge that you thereby waive your statutory right to cancel and receive a refund within this 14-day period.
2. General Policy: Outside of any statutory rights that cannot be waived, all sales are final, and no refunds will be provided once a credit has been used or the statutory cooling-off period (if applicable and not waived) has expired. Your statutory legal rights remain unaffected.

6.1. AI-Generated Content. The analysis, summaries, and the Score provided by the Service are generated by Large Language Models (LLMs), currently including OpenAI's GPT models. These models, and consequently their outputs including the Score itself, have inherent limitations, including:

1. Knowledge Cut-Off: The AI's knowledge is based on the data it was last trained on (e.g., October 2023 or June 2024, as specified in our Scope & Purpose document, Section 17). It does not have access to live, real-time internet search capabilities for generating its analysis.
2. Potential Inaccuracies & Biases: Outputs may be incomplete, factually incorrect, time-limited, contain "hallucinations" (fabricated information), or reflect biases present in the AI's training data, which may in turn lead to biased interpretations or outputs.

7.2. Personal, Non-Commercial Use. You may access and use the Service and our proprietary content (such as our analysis and Score) for your personal, non-commercial informational purposes only. This includes:

1. Reading, screenshotting, or sharing links to our public analysis pages on social media for free.
2. Quoting excerpts from our analysis, commentary, and Score in news reports, blogs, academic papers, or essays, provided you clearly attribute Checker as the source and provide a hyperlink back to the original Checker page containing our analysis.

7.3. Prohibited Uses. You may not:

1. Systematically download, scrape, copy, or store large portions of our database or our proprietary content.
2. Sell, resell, rent, lease, or commercially exploit our Service or our proprietary content.
3. Remove, obscure, or alter any copyright notices, trademarks, or other proprietary rights markings displayed on or in connection with the Service or our content.
4. Use the Service or our content for any unlawful purpose or in any manner that infringes the rights of others.

7.4. Linking & Embedding:

1. You may hyperlink to any public Checker page, provided you do so in a way that is fair and legal and does not damage our reputation or take advantage of it. You must not establish a link in such a way as to suggest any form of association, approval, or endorsement on our part where none exists.
2. If you embed our Score card or Open Graph (OG) image on another website or platform, you must not surround it with misleading context, false information, or advertisements that imply sponsorship or endorsement by Checker.
3. You must not "frame" our website or content on any other site in a way that hides our URL, branding, or the source of the content.

8.1. To the fullest extent permitted by applicable law:

1. Our aggregate liability to you arising out of or in connection with the Service, this Agreement, or your use of (or inability to use) the Service, whether in contract, tort (including negligence), breach of statutory duty, or otherwise, shall be limited to the total subscription fees you paid to us in the twelve (12) months immediately preceding the event giving rise to the claim.
2. We shall not be liable for:
   1. loss of profits, revenue, goodwill, data, or business opportunity;
   2. indirect, special, incidental, consequential, or punitive damages or loss;
   3. errors, omissions, inaccuracies, or unavailability caused by third-party platforms, services (including AI models), or data sources.

8.2. Nothing in these Terms shall limit or exclude our liability for:

1. death or personal injury caused by our negligence;
2. fraud or fraudulent misrepresentation;
3. any other liability that cannot be excluded or limited under English law.

You agree to defend, indemnify, and hold harmless nonoodle.com Ltd, its officers, directors, employees, agents, and suppliers from and against any and all claims, liabilities, damages, losses, costs, and expenses (including reasonable legal fees) arising out of or in any way connected with:

1. your access to or use of the Service;
2. your breach or alleged breach of any of these Terms, including any representations or warranties made herein;
3. your violation of any third-party right, including without limitation any intellectual property right, publicity, confidentiality, property, or privacy right arising from the links you share or your use of content obtained via the Service;
4. your violation of any applicable laws, rules, or regulations; or
5. any link to content you submit or share through the Service.

We reserve the right, at our own expense, to assume the exclusive defence and control of any matter otherwise subject to indemnification by you, in which event you will cooperate with us in asserting any available defences.

10.1. We may, in our sole discretion, immediately suspend or terminate your access to all or part of the Service, with or without notice, for any reason, including but not limited to:

1. non-payment of applicable fees;
2. breach of these Terms;
3. suspected unlawful, fraudulent, or abusive activity;
4. if required by law or ordered by a court or regulatory authority; or
5. extended periods of inactivity for free accounts.

15.4. How the automation works.

• Single, frozen prompt (as far as operationally practicable): All checks are intended to run through one core prompt structure that we publicly document in our Transparent Methodology (Section 18).
• Same treatment for every topic: The AI applies identical analytical logic to all subjects. There are no manual tweaks or different rule sets for politics, health, finance, or any other specific topic.
• Transparent sources (where possible): Each report aims to link to or describe the material the AI relied upon, so you can attempt to verify it yourself. (See Section 17.3 on citation link limitations).
• Fallibility notice: AI can make mistakes. Treat the Score as guidance, not gospel. Always exercise critical thinking.

16.2. How we're funded.
Our entire operational budget comes from the user subscription plans displayed on our Pricing page. This means:

• What we take: Monthly and annual user subscriptions.
• What we don't take: Grants, charitable donations, "philanthropic" funds, advertising, sponsorship, venture capital investment, private equity investment, or any payments tied to influencing, changing, or skipping a Score check.

17.1. What the Score is:

• AI-generated opinion: An opinion on the apparent credibility of claims made within the transcribed content of social media videos. A large-language model (currently from OpenAI's GPT series, e.g., GPT-4.1 or as updated) analyses the video's transcript, attempts to gather relevant open-source material (based on its training data), and applies a standardised public prompt (see Section 18) to produce a "Score".
• Transparent and challengeable (within limits): Each output aims to include or describe sources the AI may have relied on, allowing you to research them yourself. You can request a review if you believe an error has occurred (see Section 19 and 20).
• A cue for further scrutiny: The Score is intended to help you decide whether to trust, ignore, or further investigate a claim made in a video. It is explicitly not legal, medical, financial, or any other form of professional advice.

17.2. Key limitations you should know:

• Training Data Cut-off: The AI model's knowledge is limited to the information present in its last training dataset. For example, GPT-4.1's knowledge generally ends around June 2024 (this date may be updated by OpenAI and subsequently by us). Content or events occurring after this cut-off date may be missing from its analysis, or its assessment of newer topics may be inaccurate or incomplete. For breaking news or very recent events, always cross-check with up-to-date, reputable human-curated news sources.
• Hallucinations & Bias: Large language models can sometimes "hallucinate," meaning they may invent facts, quotes, sources, or citations that are plausible-sounding but incorrect or non-existent. They may also reflect inherent biases present in their vast training data, which can lead to skewed or unfair assessments in their output or interpretation. (Source: e.g., arXiv pre-print research on LLM limitations).
• Citation Links & Source Accuracy: AI models may struggle with generating accurate, live URLs. Therefore, instead of direct hyperlinks in some AI-generated outputs, we may provide source details (e.g., publication name, author, date) that you can use to search for the source material via a search engine. The AI's interpretation or summarisation of a source can also be flawed.

17.3. What the Score is not:

• Not lab-tested scientific proof: Like much journalistic output, the Score reflects an opinion based on information available at the time of analysis and using the specific AI model and prompt. This opinion can change if new, verified information emerges or if the analysis is re-run with different parameters or updated models.
• Not infallible: AI models are not perfect. Treat the Score as guidance, not absolute gospel. If you identify a potential error, please use the "Request a review/removal" function list on every BS Check page, or visit /contact.
• Not a judgement on motives or intent: We assess the apparent credibility of the content's claims based on the transcript, not the intentions, character, or motivations of the video's creator or speaker.

18.2. Legal & Language Guardrails (Our Aims):

1. Opinion Framing: Every analysis and Score explicitly or implicitly aims to frame its output as an opinion (e.g., by using phrases like "In our view," "appears to be," "suggests," or by the general disclaimer that it is an AI-assisted opinion). This is intended to align with legal defences such as "honest opinion" available under UK defamation law, by linking the opinion to stated or implied facts (as understood by the AI).
2. AI Disclaimer: Analyses and Scores generated by the AI model carry an explicit or implicit "[AI‑Assisted]" understanding and are subject to the site‑wide disclaimers regarding AI limitations (see Section 6 and Section 17), including the notice about potential errors and biases.
3. Safe Lexicon (Preferred Terminology): We aim to avoid high‑risk, conclusory, or overly strong labels like "liar," "fraud," "disinformation" when describing individuals or content. Preferred terms for claims lacking substantiation include "unsupported," "lacks evidence," "potentially misleading," "appears false," or "baseless," depending on the AI's assessment resulting in the Score.
4. Focus on the Claim, Not the Person: The critique is intended to be directed at the content of the claim itself, not the personal character of the speaker. Where possible and practical, particularly concerning private individuals not in public life, efforts may be made to anonymise or focus solely on the statement.
5. Right of Reply / Review: All checks, particularly those on public pages, include a "Request a review/removal" button or similar mechanism, directing users to our feedback process (see Section 19).

18.3. Corrections & Feedback System:

• A prominent "Request a review/removal" link or similar is available on analysis pages and within the service, linking to the processes detailed in Section 19 (Feedback & Complaints) and Section 20 (Corrections & Updates).
• We aim for acknowledgment of submissions and maintain a public log or updates page for significant corrections, demonstrating good‑faith efforts to correct errors. (See Section 20.5).
• Corrections, where appropriate, will generally keep the original permalink but indicate that the content has been amended, often with a note explaining the change, thereby demonstrating accountability.

18.4. Transparency Hub (This Document and Associated Information):

• This overall Terms of Service document, including this Methodology section, is linked from our website (e.g., in the footer) and is intended to be easily accessible.
• The AI model version in use (e.g., GPT-4.1) and its general knowledge cut‑off date are stated (see Section 17.2).
• We may publish an annual transparency report covering metrics such as the number of checks performed, average turnaround times (if measurable), and aggregated correction statistics.

19.1. Why we have this policy.
At Checker, we are committed to providing AI-assisted opinions and Scores that are as well-grounded as possible given the limitations of the technology and to rectifying errors promptly when they are identified. Having a clear, accessible, and easy‑to‑use feedback and complaints channel helps us:

• Maintain the accuracy and integrity of the information for our users.
• Demonstrate that we act in good faith when our analyses or Scores are challenged.
• Learn from feedback and continuously improve the Service for everyone.

19.3. What happens after you submit your request.

1. Submission Confirmation: On submission of our contact form you should receive a confirmation message on screen.
2. Human Review: Every submission is read by a human editor or designated staff member, within one (1) UK business day of receipt.
3. Triage & Logging: We will log your submission internally, assign it a unique reference ID if necessary, and categorise it (e.g., Urgent correction, Standard correction, General feedback, Complaint, Takedown request from content owner).
4. Investigation: A designated fact‑checker or editor will review the issue, examining the evidence you've provided, the original video, our analysis and Score, the sources cited (if any), and any relevant legal or policy considerations. We may contact you or the original content source for further clarification if necessary and appropriate.
5. Decision & Action: Our aim is to:
   • Address and, if verified, correct urgent factual errors that could significantly mislead readers or affect the Score, and update the public note within approximately two (2) UK business days from completion of the investigation.
   • Address and, if verified, fix standard issues (e.g., typos, broken links, minor clarifications not affecting the core Score) within approximately five (5) UK business days.
   • Respond to general feedback or complaints requiring a bespoke response within approximately ten (10) UK business days.
   • Process takedown requests from verified content owners as per Section 4.5.
6. Reply: We will email you to inform you of the outcome of our review and any action taken, unless you have requested not to be contacted or if the submission was anonymous.

19.4. How corrections appear on the site.

• Minor typographical or formatting fixes may be updated silently without a specific notification on the page.
• Substantive corrections will be flagged with a "Correction" or "Update" note, typically at the end of the article, explaining what was changed, why, and when.
• If a correction leads to a significant change in our Score, we may add a more prominent "Updated" banner at the top of the analysis page for a limited period (e.g., seven days).
• An archive or log of substantial corrections may be maintained on a public Updates & Corrections page on our website.

19.5. How we log and track feedback and complaints.

• All submissions are intended to flow into a secure internal database or tracking system that records the date of submission, the relevant article URL (if applicable), the nature/category of the feedback, its current status (e.g., Open, In Review, Resolved, Escalated), and any editor notes.
• Editors or designated staff update the status as the submission is processed.
• Senior staff may review the log periodically (e.g., weekly) to ensure items are being handled appropriately and to identify systemic issues.
• Anonymised, aggregated statistics (e.g., number of corrections made, broad categories of feedback, average response times if measurable) may be published as part of our annual transparency reporting (see Section 18.4).

19.6. Escalation Process.
If you are dissatisfied with our initial response to your feedback or complaint:

1. You may reply to the response email, clearly explaining why you remain dissatisfied and what outcome you are seeking.
2. Your concerns will typically be reviewed by a different editor or a more senior member of staff (e.g., "Head of Editorial").
3. If your issue remains unresolved to your satisfaction after this review, and involves a serious matter, you may be provided with contact details for the CEO/Founder directly in the response from the Head of Editorial (or equivalent).

20.2. How we judge the significance of an error.
We categorise errors to determine the appropriate response:

• Major Assessment Error: An error that could substantially alter a reader's understanding of the evidence supporting or refuting a claim, or that leads to a significant misapplication of our Score scale (e.g., an error causing a claimed fact to be rated with a Score indicating "well-supported" when it should indicate "unsupported," or vice-versa, typically involving a notable shift in the Score).
  • Flagging: A clear "Correction" or "Update" notice will be placed prominently, usually at the top or beginning of the analysis page, for a period (e.g. 7 days), and then potentially moved to the bottom as a permanent record.
  • Example (illustrative only): We initially assigned a Score of 3 (indicating fairly solid). Upon review, the evidence is found to be inconclusive or to contradict the claim, leading to a revised Score of 6 or more (indicating weakly supported or unsupported).
• Minor Factual Error or Slip: Typos, broken hyperlinks, incorrect dates or numbers that do not materially alter the overall assessment or the Score of the central claim(s).
  • Flagging: A concise correction note is typically added at the bottom of the analysis page. In some cases, very minor typographical errors may be corrected silently if they have no bearing on the meaning.
  • Example: We wrote "16 experts" when the referenced paper clearly lists "15 experts," but this detail does not change the overall Score or the thrust of the analysis.

20.3. What every substantive correction note includes.
For major assessment errors and significant minor factual errors, the correction note will generally state:

• What changed: The specific sentence, figure, interpretation, or Score that was updated.
• Why it changed: The reason for the correction (e.g., correction of a factual error, new information, re-evaluation based on reader feedback, misinterpretation by the AI).
• When it was corrected: The date (day-month-year) the correction was published on the site.
• (If relevant) Impact on Score: If the Score was explicitly changed, the note will reflect this (e.g., "The Score has been adjusted from X to Y as a result of this correction").

20.4. How quickly we aim to act.

• Major Assessment Errors: As soon as practicable after an editor or designated reviewer confirms the error (target: within 24-48 UK business hours of confirmation).
• All other substantive fixes: Typically within two to five (2-5) UK business hours of confirmation.

20.6. How to request a correction.

1. Click the "Request a review/removal" button found under any Score analysis, or complete the form at /contact.
2. Clearly state the URL of the Checker page you believe contains an error in its analysis or Score.
3. Describe what you think is wrong and provide any supporting evidence, sources, or reasoning you have.
4. An editor or designated staff member will review every request in accordance with our Feedback & Complaints Policy (Section 19). You can typically expect a response or see a published correction (if warranted) within the timeframes outlined in Section 19.3.

20.7. Additional safeguards relevant to Checker's AI-assisted model.

• Opinion Framing: As stated elsewhere, our analyses and Scores are presented as our AI-assisted reasoned opinion based on available evidence at the time of analysis.
• Neutral Wording (Aspiration): We strive to use neutral and descriptive language (e.g., "misleading," "unsubstantiated," "lacks evidence") rather than accusatory or definitive terms.
• No Personal Information Disclosure in Corrections: Correction notes will not reveal the personal identity of the individual who submitted the content link for analysis or who reported the error, unless they have expressly consented to be named (which is not our standard practice).

In short:

• We aim to correct significant assessment errors, including those affecting the Score, quickly and transparently.
• Major changes get a prominent notice; minor factual corrections get a footnote or end-note.
• Every substantive correction note explains what changed, why, when, and any impact on an explicit Score.
• Significant corrections are intended to be logged publicly.
• Anyone can flag a potential issue – use the "Request a review/removal" button or email us, and we will investigate.