How the Pinterest Algorithm Works: Find, Score, Rank, Mix

Pinterest's own research papers describe four steps: find, score, rank and mix. What each step reads, and what you control at each one.

Quick answer: The Pinterest algorithm isn't one secret score. Pinterest's research describes how its systems retrieve candidate Pins, score search relevance, rank personalised recommendations and diversify the Home Feed. Find, score, rank and mix is a useful way to understand those jobs, but they're not one identical four-step pipeline across every Pinterest surface.

You can't control who is scrolling. You can control what Pinterest knows about your Pin at each step.

Where This Comes From

Everything here is from Pinterest's own publications: research papers by Pinterest authors, posts from Pinterest Engineering, and the Pinterest Help Center and Newsroom. They describe systems Pinterest has built and deployed across different surfaces, including Search and Home Feed. The four steps below are our explanatory framework, not Pinterest's official name for a single algorithm. The papers don't publish a universal set of signal weights, so be wary of anyone who says titles count for a fixed share or saves are worth so many clicks.

Step	What Pinterest does	What you control
1. Find	Pulls a shortlist of Pins that match what someone is into	Clear boards, an on-topic image
2. Score	Grades how well each Pin matches a search	A picture that shows the topic, the same words in the title
3. Rank	Predicts what this person will save or click	Timing
4. Mix	Spreads look-alike Pins apart	Pins that are really different

Step 1: Find

Pinterest has billions of Pins. Before anything is ranked, a fast first step pulls a shortlist of candidates for each person. Pinterest describes its home feed and search as working like a funnel, wide at the top and narrowing down.

To do that, Pinterest has to know what your Pin is. Its published models learn that from the picture, the words, and the boards the Pin is saved to.

Why Boards Matter

Pinterest's PinSage model learned what a Pin is from the boards people save it to and what else sits on those boards. Search boards for "chic winter outfits" and you get dozens with exactly that name. That is the kind of signal it learns from.

Save your Pins to boards that are clearly about the topic, with plain names. For a starting point, the free Board Name Generator suggests board names from your topic that you can edit and copy into Pinterest. For more on naming, see Pinterest Board Names.

Step 2: Score

When someone searches, Pinterest grades each Pin on how well it matches the query. Its search paper describes a relevance model that rates Pins on a five-point scale, from highly relevant to highly irrelevant.

What the Model Reads

The same paper lists the text it reads:

Your Pin title and description.
The names of the boards it is saved to.
The title and description of the page it links to.
Searches where the Pin did well before.
A caption that an AI writes of your image.

That last one surprises people. Pinterest describes your picture to itself, then checks that description against the search.

So make sure the picture shows what the Pin is about, and use the same words in the title. A Pin titled "chic winter outfits" whose photo is a close-up of a handbag is making two different promises.

Step 3: Rank

Now Pinterest guesses what this particular person will save or click. Its published ranking model reads what they did just minutes ago, their recent actions in sequence.

You can't control someone's recent activity. You can choose when your Pins are published, although the research does not establish publishing time as a direct ranking signal.

In Pinterest Trends, "chic winter outfits" in the US climbed from October last season and peaked in early December. In early October this year it was starting to rise again. A Pin that is live while people are actively looking for the topic has more of those people to be shown to. Check your keyword's curve and have your Pins up before the climb.

Step 4: Mix

Before the feed shows up, a last pass spreads look-alike Pins apart. Pinterest Engineering has described comparing Pins by how they look, what they say and how they connect, then spacing similar ones out so the feed stays varied.

Similar-looking Pins can be spaced apart in the Home Feed's final re-ranking. That doesn't establish a fixed limit on how many Pins from one creator can appear, but it does make near-duplicate designs a weaker bet for feed exposure.

Make each Pin really different: a new picture, layout or headline, not just new words on the same design. If you make Pins in PinEasy, Fresh Pack checks every Pin in a pack for this.

The Viewer's Own Dial: AI Content

Since October 2025, people on Pinterest can choose to see less AI-generated content on eligible image Pins, category by category. Pinterest labels AI-modified images using image metadata, its own classifiers and creator disclosure.

Pinterest presents this as a choice for viewers, not a penalty for creators, and hasn't said anything about reach. If your image is AI-made, say so. Where it matters to people, use real photos.

A Pin Checklist for the Algorithm

Clear board? The Pin is saved to a board that is clearly about its topic.
Picture matches? The image shows the topic, and the title uses the same words.
Page delivers? The link goes to a page that keeps the Pin's promise.
Timing? You checked the keyword's curve and published before it climbs.
Really different? This Pin isn't a near-copy of your last one.
AI transparency? Check Pinterest's AI labelling rules, and disclose AI use where applicable.

For where to put keywords and how to measure results, see Pinterest SEO.

Frequently Asked Questions

How does the Pinterest algorithm work?

Pinterest uses different systems to retrieve Pins, score search relevance, rank personalised recommendations and diversify the Home Feed. Find, score, rank and mix is a useful summary of those functions, not one official four-step sequence used everywhere.

Do boards matter for the Pinterest algorithm?

Yes. Pinterest's published models learn what a Pin is partly from the boards it's saved to, and board names are among the text its search model reads.

Does Pinterest use hashtags to rank Pins?

Hashtags aren't among the inputs Pinterest lists in its published search research. Put the words in your title, description and board names instead.

Does Pinterest penalise AI-generated Pins?

Pinterest hasn't said it does. It labels AI-modified images and lets viewers choose to see less AI content in some categories. That's a viewer setting, not a stated creator penalty.