GoodRepHow Review Algorithms Work on Google, Yelp, and Facebook
The algorithms are different in detail but similar in shape: each platform weighs reviews on signals that aren't shown publicly, and each one filters out reviews that look suspicious. Knowing the patterns is the difference between a profile that performs and one that quietly underperforms.
Most local business owners think about reviews in terms of star count and rating average. The platforms think about reviews very differently. Google has a local ranking algorithm that considers reviews as one input among many. Yelp has a controversial filter that hides "not recommended" reviews. Facebook treats recommendations more like social posts than ratings. Each platform is doing its own math, and the math affects what customers see.
This post is the working model of how the three major platforms handle reviews behind the scenes, what's known versus inferred, and what it means for how you collect and respond.
Key takeaways
- Google weighs review count, recency, response rate, keywords, and rating in its local pack ranking; each one is partial.
- Yelp's filter hides up to 70% of submitted reviews on some profiles; most of what's hidden is from new accounts.
- Facebook recommendations behave more like social content; engagement matters more than star ratings.
- No platform publishes its full algorithm, but the patterns are observable.
- Working with the algorithm mostly means doing the same things compliant review collection requires anyway.
Google: Local Ranking and Review Signals
Google's local pack (the three-business map result) is the most valuable digital placement most local businesses can have. Google has confirmed that reviews are part of the ranking, and over the years they've published enough hints that the working model is fairly clear.
The signals Google appears to weigh:
Review count. Total number of reviews. More reviews suggest established prominence. The relationship isn't linear; the gap between 5 and 50 reviews matters more than the gap between 200 and 800.
Review recency. How recently new reviews have arrived. A profile with reviews in the last 30 days outranks an equivalent one with reviews from 12 months ago.
Owner response rate. What percentage of reviews have an owner response, and how quickly. Google has stated that response behavior is one of the activities that influences local rankings.
Review keywords. The actual words customers use. A plumber whose reviews mention "emergency repair" repeatedly may rank better for "emergency plumber" searches in their area. This is also why response keywords matter: when you respond mentioning a service or location, you reinforce relevance.
Average rating. The visible star number. This is part of the signal but less dominant than most owners assume; a 4.6 with 80 reviews and high recency outranks a 4.9 with 8 stale reviews regularly.
What Google doesn't seem to weight as much: review length, photo content in reviews (some signal but small), or reviewer identity (although the trust score of the reviewer affects whether the review counts at all). Reviews and local SEO covers the integration with the broader ranking system.
The practical implication: Google rewards consistency more than perfection. A steady stream of moderate-rating reviews from real customers, each responded to, beats a burst of perfect reviews followed by silence.
Google's Spam Filter
Google has a spam filter that removes reviews it considers fake, incentivized, or otherwise non-compliant. The filter is opaque, but the patterns of what gets removed are observable:
Reviews from accounts with no other activity (single-review accounts) are often filtered. Reviews posted in suspicious bursts (50 reviews in a week from a previously slow profile) often get removed. Reviews that contain keywords associated with paid promotion ("amazing service" in suspicious patterns) are flagged. Reviews from accounts whose IP addresses geographically don't match the business location can be flagged.
The filter sometimes removes legitimate reviews, which is frustrating. But the underlying pattern is consistent: reviews that look organic survive, and reviews that look like a campaign get filtered. This is one reason burst review pushes don't produce the count gain owners expect.
How to get more customer reviews compliantly covers the line between what survives the filter and what doesn't.
Yelp: The Filter That Hides Most Reviews
Yelp's algorithm is the most aggressive of the three. Yelp uses what it calls a "recommendation software" that distinguishes between recommended reviews (visible by default) and not-recommended reviews (hidden behind a small link at the bottom of the page).
For many businesses, the filter hides a striking percentage of submitted reviews. Some sources have reported up to 70% on certain profiles. The filtered reviews still exist (you can see them by clicking "not recommended reviews") but they don't count toward the visible star average.
The pattern of what Yelp filters:
Reviews from new accounts (no profile picture, no other reviews, no friends) are heavily filtered. Reviews from accounts that have only ever reviewed one business are filtered. Reviews that look like they were prompted in any way (similar wording across several reviews of the same business) are filtered. Reviews from far-away IPs are filtered. Reviews from established Yelp users with active profiles are almost always recommended.
The implication is that Yelp rewards established Yelpers over your customers who happen to be on Yelp. A loyal customer who has used Yelp for years and posts reviews regularly will land. A customer who creates an account specifically to leave you a review usually won't.
This is also why Yelp asks businesses not to ask customers for Yelp reviews. The platform considers solicited reviews suspicious by design. Yelp optimization for local businesses covers the practical workarounds.
Facebook: Recommendations as Social Content
Facebook moved away from traditional star ratings several years ago in favor of a recommendation model: customers can recommend a business, optionally with a written explanation, and the business shows a count of recommendations rather than an average rating.
Behind the scenes, Facebook treats recommendations more like social content than like reviews. Engagement matters: comments, reactions, shares, and the network effect of friends seeing each other's recommendations all influence reach.
The algorithm signals appear to include:
Whether the recommendation has text and photos (more engaging recommendations surface higher). Whether the recommender's friends interact with it (network effect). How recently the recommendation was posted. Whether the business engages with the recommendation (a reply from the business surfaces it more).
Facebook also has a spam filter, but it's less aggressive than Yelp's and less opaque than Google's. Most authentic recommendations from real accounts survive.
The practical implication: Facebook rewards engagement and social authenticity. A recommendation that includes a photo and gets a few comments will reach further than a one-line text recommendation, even if both are technically positive. Facebook reviews and recommendations covers the platform's specifics.
What All Three Platforms Have in Common
The platform algorithms differ in detail but share several patterns worth knowing:
Authentic over orchestrated. Every platform's filter is trying to detect orchestration. Reviews that look like they came from real customers responding to real experiences survive. Reviews that look like a coordinated effort don't.
Established over new. Every platform weights reviews from established accounts higher than reviews from brand-new accounts. The implication: customers who already use the platform are higher-leverage than customers you're convincing to create an account.
Engagement over silence. Owner responses are weighted positively across all three platforms. Whether the platform calls it "ranking factor" or "engagement signal," the effect is similar.
Consistency over bursts. All three platforms detect burst patterns and treat them as signals of orchestration. Steady is better than bursty.
The consistent message across the three: do compliant, organic, consistent review work, and the algorithms reward it. Try to game it with incentives, gating, or burst campaigns, and the algorithms penalize it.
What This Means for How You Work
The algorithm patterns suggest a few practical priorities:
Prioritize Google for the ask. It's the platform with the largest impact and the cleanest filter for legitimate solicited reviews.
Don't ask for Yelp reviews directly. Yelp's policy and filter make solicited reviews counterproductive. Customers who use Yelp will leave Yelp reviews on their own. Make sure your Yelp profile is claimed and current so the organic reviews land somewhere appropriate.
Treat Facebook as part of the social-plus-review work. Engaging with recommendations matters as much as collecting them. Photos and replies amplify reach in a way that doesn't apply to Google or Yelp.
Respond consistently across all three. Every platform rewards response behavior. The cost of responding is small. The compounding effect across all three platforms is meaningful.
Don't run burst campaigns. Steady drip beats bursts on all three platforms, both in terms of what the algorithm rewards and what customers reading the profile perceive. Review velocity covers why steady flow outperforms.
What's Changing
The platforms update their algorithms regularly. The specifics shift, but the underlying principles have been stable: reward authenticity, penalize orchestration, weight engagement, prefer recency. Anyone who claims to know the exact current algorithm is overstating; anyone who claims the principles are stable is closer to right.
A few directions worth watching: AI-generated reviews are an emerging detection target across all platforms; Google has been tightening enforcement on incentivized reviews; Yelp's filter has if anything gotten more aggressive over time; Facebook's algorithm continues to shift toward favoring recommendations with engagement.
The defensible practice is to follow the underlying principles regardless of the algorithm specifics. Authentic, consistent, responsive review work is the baseline that ages well.
The Bottom Line
The three major platforms each have their own math, but the practical patterns rhyme. Google rewards a balanced profile of count, recency, response rate, and rating. Yelp's filter heavily favors established users over solicited reviews. Facebook treats recommendations more like social content than like ratings. Across all three, the algorithm rewards the same behavior: authentic reviews from real customers, steady volume rather than bursts, and consistent owner engagement.
You don't need to game the algorithms. You need to do the work the algorithms reward.
GoodRep handles the consistent practice that the algorithms reward across all three platforms: steady asks, single-inbox responses, and visibility into the signals each platform weighs. $39/month, 14-day free trial. Start free.