Last updated September 15, 2026
nextbest is community-sourced skincare product rankings. We aggregate real discussions from Reddit skincare communities to show you what people actually use and recommend. We also grade the peer-reviewed studies on skincare active ingredients, starting with acne, so you can see what the research says alongside what people report. No influencers, no paid placements — just real people sharing what works for them.
We scan thousands of Reddit threads from skincare communities. AI reads through them and picks out product mentions, noting what people recommend and why. Separately, we search PubMed for peer-reviewed studies of active ingredients and grade how well the research supports each one for a concern — see how we grade research. On most concern pages, products are ranked by a community score — how often the community mentions them, weighted by how recent each mention is, how upvoted it was, and which subreddit it came from. The acne page blends that community score with research evidence on each product's active ingredients. Everywhere else — and on the handful of concerns with too few products to rank meaningfully — it's simple mention volume: more mentions means a higher ranking.
nextscore is a single 0–100 number for how well a product has been received. It is calculated from verdicts — mentions where someone came down clearly on one side, saying a product worked for them or that it didn't. nextscore is based on the share of a product's verdicts that are positive, adjusted for how much evidence sits behind it. Today those verdicts come from community discussion on Reddit: the eleven subreddits listed below. Peer-reviewed studies are graded separately and are not part of nextscore. nextscore is defined over verdicts, not over the source those verdicts come from.
Not every mention is a verdict. Mentions that are neutral or mixed — “it's fine, I guess”, “lovely texture, but it broke me out” — are counted as discussion, not as verdicts, and are excluded from the score. They tell you a product is being talked about; they don't tell you whether it worked. Scoring them somewhere in the middle would invent an opinion nobody expressed — a passing mention is not a review.
A product with few verdicts is pulled toward the catalog-wide average, and only earns a distinctive score once it has enough evidence behind it to pull away. That pull is worth 10 verdicts, which is why five positive verdicts out of five reads as well-liked but thinly evidenced, rather than as a perfect 100. The pull works the same way in the other direction: five negative verdicts out of five land in the mid-50s rather than near zero. A middling score usually means there isn't much evidence behind it yet — not that a product is mediocre — so read it alongside the verdict count. Below 5 verdicts we show no score at all — just “not enough verdicts yet”.
Every score is shown with the number of verdicts behind it, because the count is what tells you how much to trust the number. Scores update as new discussion is ingested; the pipeline runs daily, so a score moves as the conversation does. It is a read on reported experience, not a lab test or a verified-purchase review.
The eleven communities we source are chosen for experience-based discussion — people reporting what a product did on their own skin over weeks and months — not for volume. A bigger, more general forum would produce more mentions and fewer verdicts worth counting. The full list is below.
Alongside community discussion, nextbest grades the peer-reviewed research on skincare active ingredients for specific concerns, starting with acne. This section explains where the studies come from, how each one is read, what each grade means, and how grades affect the acne ranking.
Our research source is PubMed, the index of biomedical literature run by the U.S. National Library of Medicine. We search it separately for each ingredient and concern, asking PubMed for meta-analyses, systematic reviews, randomized controlled trials and other clinical trials, and retrieve up to 40 matching records per search. A study found by an earlier search is kept even when a later search no longer returns it.
Each study's design comes from PubMed's own publication-type labels, not from AI. Most of the studies we collect are meta-analyses, systematic reviews, randomized controlled trials or other clinical trials; a study PubMed labels as another design, such as an observational study, keeps the label PubMed gave it.
An AI model reads each study's abstract and records whether it is relevant to the ingredient and concern, whether it measured the ingredient itself, which way the result went, and any certainty rating the paper gives its own evidence. Only studies that measured the ingredient itself count toward a grade; a review that lists the ingredient among many treatments without reporting on it does not.
We show a grade for an ingredient and concern only when there are at least 2 relevant studies, including at least 1 randomized controlled trial or a stronger design such as a meta-analysis or systematic review. Each grade means something specific:
More than half of the studies that measured the ingredient found a benefit, across at least 10 such studies, including at least 1 positive meta-analysis or systematic review.
Strong evidence means more than half of the studies that measured the ingredient itself found a benefit, there are at least 10 studies that measured the ingredient itself, and the ingredient has at least 1 positive meta-analysis or systematic review that measured it and whose authors do not rate their evidence low or very low. If any positive meta-analysis or systematic review that measured the ingredient itself rates its own evidence low or very low, the grade can be no higher than Some evidence.
More than half of the studies that measured the ingredient found a benefit, including at least 2 positive randomized controlled trials or stronger designs.
Some evidence means more than half of the studies that measured the ingredient itself found a benefit, and at least 2 positive randomized controlled trials or stronger designs are among them, without meeting the bar for Strong evidence.
The research doesn't establish a benefit yet; this grade is not a warning that an ingredient is harmful.
Not proven yet covers everything else, however many studies there are: half or fewer of the studies that measured the ingredient itself found a benefit, no study measured the ingredient itself, or too few positive trials or reviews measured it to reach Some evidence. Not proven yet means the research doesn't establish a benefit yet, not that an ingredient is harmful.
On the acne page, a product's ranking blends 75% community score with 25% research evidence. A product's evidence value is the best grade we show for any of its verified active ingredients, counting only actives named in the product's name, given a stated strength, or listed high in its ingredient list. A product with no graded active ingredient gets a neutral evidence value, the same as a “Not proven yet” grade, so a product is never marked down for research that is missing or thin. It can still be passed by a product whose actives have stronger evidence. Only Some evidence or Strong evidence raises a product's score above that neutral value.
Research grades are separate from nextscore: studies are not part of nextscore.
For an example, see the studies on Adapalene for acne.
Research context, not treatment advice. Studies are about the ingredient, not any one product.
We pull from eleven subreddits, chosen for their high-quality, experience-based discussions:
Rankings are based on genuine community signal — a recency- and upvote-weighted community score on concern pages (blended with graded research evidence on the acne page), mention volume elsewhere — never on paid placements. We don't accept money from brands to boost rankings. Affiliate links, where present, don't influence which products appear or where they rank. The methodology is transparent: you can see which threads informed each product's ranking, so you can verify it yourself.
Questions? Reach out to kayleigh@nextbest.one.
We use cookies to improve your experience. Learn more