ScrollSensay

Framing the Problem

The current browsing experience doesn’t help users verify information online in the moment:

• leading to losses from scams and misleading claims

• sharing (thinking it helps others)

• feeling anxious or confused

One of the benefits of our solution is that older adults are disproportionately affected by misinformation.

Helping them first creates a huge social impact.

Idea Explanation

ScrollSensay:

• Works as a Chrome extension

• Adds a simple badge to the bottom-right on the screen

  • red

  • amber

  • green

• Explains in simple terms why information might be worth a closer look before clicking, trusting, or sharing a post.

Implementation

ScrollSensay is a Chrome extension that works with a Python backend which uses Gemini. The frontend and backend communicate to analyse posts which then returns a transparency score and an explanation.

On the frontend, the browser extension:

• runs in JavaScript

• monitors the user’s Facebook feed as they scroll

Whenever the user stops scrolling for a period of time and text is visible, the extension extracts the text directly from the page. The extracted content is then sent to our backend for analysis.

The backend uses a fact validation prompt and sends it to Gemini (gemini-2.5-flash) via API. Gemini identifies the main factual claim in the post, checks its authenticity and produces a structured judgement that includes:

• a score from 0.0–1.0, which the extension maps to green (≥0.7), amber (0.4–0.69), or red (<0.4).

• a short explanation

Once the backend returns this result, the extension immediately reflects the colour and emoji in the bottom-right badge. Users can click the badge to view the explanation or choose an audio readout for accessibility.

Challenges

Challenge: Sounding Patronising

Response: We use neutral, supportive wording e.g. "check before clicking," not "you are wrong."

Risk: Labelling Opinions as False

Response: We separate factual claims from opinions and do not rate them using the RAG method.

Challenge: False Confidence

Response: We avoid "87% fake" style outputs and use clear RAG transparency labels.

Challenge: Mobile Platform Limits

Response: The MVP cannot run directly over the mobile apps because they do not allow browser extensions to overlay. We therefore focused on a Chrome desktop extension.

Accomplishments

• Through effective team collaboration, we went beyond creating prototypes by delivering a working extension capable of recognising misinformation, humour / political satire while scrolling to promote transparency without being intrusive

• Our solution can be scaled across social media platforms, news sites and blogs so ScrollSensay has global potential.

• Designed an accessible user experience requiring minimal interaction. Users can enable or disable Sensay's Advice with a single click, making it easy for anyone to use regardless of technical ability.

• Created more than a traditional fact checking tool. Sensay encourages users to think critically as they browse by providing contextual advice, highlighting patterns of potentially misleading content and helping raise awareness of online scams and misinformation.

Next Steps

• Expand analysis beyond written text by incorporating image analysis to identify potentially AI-generated images and explain why the image seems unauthentic.

• Broaden content analysis to include comments, replies and other post interactions, providing users with more comprehensive insights.

• ScrollSensay uses AI for an effective and scalable system however, its evaluations are not always correct. This can be improved by introducing a verified user feedback system, allowing users to report incorrect evaluations with reasons.