NewsScore

The project helps channel editors and administrators sort through subscriber news submissions. The bot analyzes incoming materials, evaluates them against several criteria, and generates a prioritized queue. The core of the solution involves assessing the timeliness of the submission, signs of urgency, available source information, and its relation to other materials. Editors can see which news items require immediate attention and can verify the rationale behind the bot’s recommendations.

Administrator or editor of a mid-to-large news Telegram channel receives a daily stream of content from subscribers: photos, videos, and rumors. Everything arrives chronologically, not by importance; urgent, duplicate, or reliable content must be manually sifted through. This problem results in wasted time, vital information getting lost in the feed, and inaccurate news potentially going live.

Administrators and editors of Telegram channels covering news, local updates, and specific topics, which receive a regular flow of reader-submitted content.

Segment: Local news channels. In our study, representatives from these channels reported recurring messages about the same events, peak traffic load, and instances where certain topics or clarifications were lost.

News and information channels are currently topping popularity charts. Telegram itself is integrating AI and automation for incoming bot messages—the market is clearly moving in this direction.

In this research, I compared three areas: local news and resident feedback, pop culture, beauty, and lifestyle. In local channels, rapid response, cross-referencing testimonies, and managing follow-ups are crucial. In pop culture, the original source, date, and accuracy of translation play a significant role. In the other cases, the primary stream consisted mostly of invitations and proposals, which is less aligned with the original NewsScore objective. For the initial review, I am selecting local news channels for three reasons: a confirmed flow of materials; the need to correlate messages; and specific instances of information loss.

Subscription ≈ 500 ₽ per month. The channel admin pays for it; the bot saves him time on analysis. An interview showed that TGK is willing to pay for an additional tool.

1. Admins from TGStat and Telemetr catalogs2. SMM communities and channels owners' chats. 0 ₽ in money, the price is time and reputation3. Recommendations from early users within the admin circle. 0 ₽, but depends on the quality of the pilotLTV ≈ 3,000 ₽ = 500 ₽ × 6 months. Assumption: the client stays for six months, requires verification.

The core idea of the project remains the same: NewsScore helps the editor score incoming materials using intelligent scoring. The research refined the requirements for this system.

We assumed: the identified niche was virtually unoccupied. We found out: ready-made bots, builders, and established manual processes exist. We now believe: the product must demonstrate its advantages over specific alternatives.

We assumed: repetition primarily impedes analysis. We found out: R3–R5 describe valuable new content embedded within repetitive messages. We now believe: scoring must account for the value of the addition, while grouping must preserve the original sources.

We assumed: the sender’s history allows for reliability assessment. We found out: a familiar author can also be mistaken, and a new author can submit quality content. We now believe: the author’s history is a supplementary attribute, separate from fact-checking.

We assumed: a single prioritization order suits all channels. We found out: editorial needs and attitudes toward sorting vary. We now believe: criteria must be validated considering the subject matter and the editorial workflow.

We assumed: low subscriptions and a single administrator would ensure simple integration. We found out: team consensus and differing payment preferences occur. We now believe: the purchasing path depends on the user’s role and the proven utility.

Problem—confirmed. Editors manually match incoming items, determine priorities, and may miss topics or clarifications. Basis: individual scenarios R1, R3–R6.

Segment—confirmed. A workflow exists among local news channels that feature regular reader-submitted material. Basis: R1, R3, R6. The size of the paying segment is currently unknown.

Unique Value Proposition—unverified. An explainable news queue that helps the editor quickly identify what requires attention.

Solution—unverified. Material intake, scoring by relevance and urgency, source and supplementary information tracking, grouping, explanations, and manual correction. The effectiveness of this combination requires testing.

Channels—unverified. Direct outreach, specialized communities, and recommendations. Actual attraction effectiveness has not been measured.

Revenue—Insufficient data. The primary projected revenue stream is a monthly subscription from the channel owner or editorial team. The estimated price of 500 rubles has not been confirmed by payment. R1 and R5 provide conditional price benchmarks.

Costs—Confirmed. For a small pilot, I am preliminarily budgeting 3–9 thousand rubles per month: 2–4 thousand for the server, database, and storage, and 1–5 thousand for AI text analysis. I will handle support independently and plan to attract initial users through direct outreach without an advertising budget.

Metrics—Unverified. For the first pilots, five key indicators were selected: number of connected channels; retention after the trial period; active time spent on analysis; number of significant materials missed; conversion from pilot to paid subscription. Grouping errors and reasons for manual score adjustments will be used as additional quality metrics.

Unfair advantage—Insufficient data. A protected advantage has not yet been established. The research helps design the product more accurately, but by itself, it does not prevent copying.