Why Crowd-Sourced Ratings Work (When You Defend Them Right)

"Open ratings always get brigaded into uselessness." That is the standard objection to a site like Internet Standing, where the public rates the people it follows — creators, journalists, commentators, public figures — and the number on each record, their Standing, is nothing more than those ratings, weighted and mapped to 0–100. The objection is real, and it is solvable. Here is the case for crowd-sourced ratings done right: the three ways a naive version fails, and the defenses that hold.

Why a panel has a ceiling

Start with what the public's ratings are competing against: a small panel of reviewers who decide what each record says. The panel model has real strengths:

• Its method is written down and applied the same way to everyone.
• Reviewers can be vetted.
• Bad-faith reviewers never get a vote.
• A dispute has one accountable team to take it to.

And one structural weakness:

• It cannot scale. A panel rates hundreds of people a year, perhaps thousands. The internet has millions of accounts with real audiences — on YouTube, Instagram, TikTok, X, Twitch, Telegram and in the newsrooms. The long tail is invisible to a panel.

The naive crowd-sourcing failure modes

Just opening up "rate this person" to the public produces three predictable failures:

1. Brigading

A coordinated group can dogpile anyone. If the loudest five percent of a creator's detractors all leave one-star ratings on day one, the public number moves before the quieter majority of their audience has had a chance to balance it. This is the canonical objection.

2. Self-selection

People who are motivated to rate someone skew toward strong feelings. Casual viewers do not bother. The pool of raters is not the audience; it is the loudest fraction of the audience.

3. Sock puppets

Someone with a grudge opens a hundred accounts and rates from all of them. Without strong identity signals, this attack is essentially free.

Naive crowd-sourced sites show all three failures. That is why they have a bad reputation. Solving them takes deliberate design.

The three defenses that actually work

1. Weighted ratings (against brigading and sock puppets)

Not every rating should count equally. A rater's weight scales with their own record:

• Have they rated people across the range of what they watch, or only ever in one direction?
• Do other raters find their ratings helpful?
• How long has the account existed, and does it show signs of organic use?
• Has it been flagged for spam or coordinated voting before?

A rater who scores well on these carries more weight than a brand-new account with no history. Every rater can see their own weight, and a record prints it beside each rating. A coordinated raid then needs coordinated histories, which means months of patient sleeper accounts that are mostly not worth maintaining.

2. Showing the footing, not just the number (against self-selection)

A Standing is never published alone. The record shows how many ratings it rests on, the words raters used, and the line of the last thirty days. A person whose audience is split reads as split; a sudden cluster of new ratings after one story shows up as a jump on the line, and you can judge it for yourself rather than take a single figure on trust.

3. Vote-ring detection and a publication quorum (against pile-ons)

Ratings that look coordinated — a ring of accounts rating the same people the same way — are kept out of the number. And no Standing is published at all until a record has 20 ratings; until then it says "No standing yet" and counts up, so a handful of early ratings cannot stamp a number on anyone. This means:

• A sustained shift in opinion moves the Standing as independent ratings accumulate.
• A burst of look-alike ratings is filtered rather than averaged in.
• Each rating fades gently with age, so the record follows the person as they are now.

The trade-off: a young record says "No standing yet" for longer. We think that is the right trade-off — a number built on too few ratings is noise, or an attack.

Why this does not collapse into "a trusted authority decides"

Critics sometimes argue that any moderation amounts to recreating the panel. There is a meaningful difference:

• A panel decides what the rating is. The reviewers are the source of truth.
• Weighting decides whose ratings count more. The public is still the source of truth; the site only discounts low-quality contributions.

A rater who consistently files thoughtful ratings across the range of what they follow has earned a higher weight. One who only ever leaves one-star or five-star one-liners has not. That is not the same as a panel deciding who is reliable. The mechanism is earned, not appointed.

How a defended crowd beats a panel at scale

Once the defenses are in place, crowd-sourced ratings have a structural advantage a panel cannot match:

1. Coverage. Anyone the public takes the time to rate gets a record. The long tail is reachable.
2. Currency.Ratings are continuous. A Standing reflects this month's audience, not a review written years ago.
3. Visible footing. Every number comes with its count, its movement and the words behind it.
4. Accountability of the rater. Raters on Internet Standing have a profile and a weight anyone can see. Panel reviewers usually do not.
5. Cost. Free for readers and raters, with no paywall in front of the record.

What crowd-sourced ratings still cannot do

Honest about the gaps:

• Rate a single video or post at scale.Panels and crowds both struggle here. A record is a person or an organisation, and one post is read on its account's record.
• Rate someone new quickly. A record needs 20 ratings before it shows a Standing. A panel can rate a newcomer on day one.
• Split by topic.No rating splits a person's accuracy by subject. That is an open problem.

The synthesis

Crowd-sourced ratings work — but only when the defenses are explicit. A site that pretends brigading is not a problem, or hides the count behind a single number, is reproducing the failure modes critics rightly point at.

A site that documents its weighting, publishes the count and the movement with every number, and filters coordinated ratings has a real shot at being more useful than a panel at scale — while being honest about the failure modes it has not fully solved.

That is the bet Internet Standing is making. The methodology page documents how each defense works in detail, and How it works walks through a record from its first rating to a published Standing.

Related

• Internet Standing methodology
• How to verify news credibility in 2026
• Political bias in Indian YouTube: three ecosystems, one feed

Have a critique of the method we haven't addressed? Email methodology@web-jury.com. Genuinely interested in good-faith pushback.