Turn your content into Pinterest growth.

Pinly reads your site. Designs the Pins. Publishes them. Then tells you which ones worked, so the next batch is better than the last.

Free to start. Connect Pinterest when you’re ready to publish.

Why Pinterest

Not because it’s trendy. Because it works differently from every other platform, and that difference favours small teams who show up consistently.

How long one post stays alive

Illustrative — the shape, not real data
A post in a feedA Pinpublishedweek 1month 3month 12

This isn’t luck. A feed sorts by what is new, so attention spikes and dies. Search sorts by what is relevant, so attention builds and stays. Everything else on this page comes from that one difference.

People show up looking for something

Pinterest works like a search engine. Someone types what they want. You show up as the answer. You’re not interrupting anyone — they asked.

Your Pin gets filed, not buried

A feed post is ranked by how new it is. In a day, it’s gone. A Pin gets filed under what it’s about, and keeps getting found months later.

Every Pin sends people to you

The link goes to your site. Pinterest is one of the only big platforms where sending people away is the point, not something the algorithm punishes.

The work stacks up

Old Pins keep getting found. What you posted in March is still working in November. That changes the math on posting regularly — and it’s why showing up beats going viral once.

The catch: this only pays off if you keep posting, and you only get better if you know which Pins actually worked. Both are boring. Both never stop. That’s the part Pinly takes off your plate.

How it works

Three steps. Each one feeds the next.

  1. 1

    Generate

    Pinly reads your site, works out what you actually sell, and hands you Pin ideas built from your own pages and products.

    your site
    5 matcha rituals for slow morningsMatcha vs. coffee, honestlyA beginner’s matcha kit
  2. 2

    Publish

    Pick the ones you like. Drop them on a day. Pinly posts them through the official Pinterest API. Close your laptop — they still go out.

    M
    T
    W
    T
    F
    S
    S

    Queued on our servers. Retries if it fails. Can’t post twice.

  3. 3

    Learn

    Every Pin remembers how it was built. So you get results by headline style, angle, template and board — not one number that tells you nothing.

    How-toListicleQuestionQuote

    Split by headline style. The fourth has too little data to call.

The premise

Making the Pins is the easy part

Point AI at your website and you’ll have fifty Pins by lunch. They’ll look fine. They’ll also look exactly like what your competitor gets from the same prompt. AI gets cheaper every month. Knowing which of your Pins actually worked doesn’t.

The shortcut

what most tools sell you
  1. Your website
  2. One prompt
  3. 50 Pins

Fifty Pins. One decision behind all of them. Nothing measured, nothing learned — so batch fifty-one is the same guess as batch one.

The system

what Pinly is for
  1. Real content
  2. Real demand signals
  3. Controlled variants
  4. Real performance
  5. A better next batch

Every Pin is a test, and Pinly remembers how you built it. Three of these beat your average — and you can see exactly what they had in common.

Here is the whole idea. AI is something you rent. The link between what you chose → who saw it → what happened is something you own. And it’s worth more every month you keep going.

The architecture

Three things Pinly knows before it writes a word

What you sell. What people are searching for. What already worked on your account. Only then does it make anything. Most tools start at step two and skip the rest.

Business knowledge

  • Products
  • Website & blog
  • Brand kit
  • Real imagery
  • Audience

Market signals

  • Keyword clusters
  • Trend direction
  • Seasonality
  • Search intent

Performance memory

  • Every past Pin
  • Its full design record
  • What beat the average
  • Confidence
  1. 1Strategy engine

    Built today

    Picks what to make and why. Which topic, which angle, which product — and what you haven’t covered yet.

  2. 2Creative engine

    Built today

    Turns that into a brief, then into real Pins. A template, your own photo, copy written for the angle, and versions that actually differ.

  3. 3Quality gate

    Built today

    Blocks anything with cut-off text, weak contrast, filler copy, stuffed keywords, or a claim your site doesn’t back up.

  4. 4Publish

    Built today

    Goes into a queue on our servers, retries if Pinterest hiccups, and posts through the official API at the time you picked.

  5. 5Measure

    Built today

    Pulls the numbers daily and ties them back to every choice that went into the Pin. Not one anonymous total.

  6. 6Learn

    Built today

    Works out what actually beat your average, across how many Pins, and whether that’s enough to trust yet.

  7. 7Next experiment

    Built today

    Change one thing on purpose. Hold everything else. Now the next result actually tells you something.

Then back to the top, smarter than last time. Every cycle adds to what Pinly knows about your account, and that’s the third input it reads before it makes anything. It’s the one part nobody can copy.

Read it backwards and it explains itself. You can’t run a test without a record. You can’t have a record unless the Pin was built from a structure. And you can’t structure anything until you know the business.

Imagery

Three layers. Only one of them is AI.

“Generate a Pinterest Pin” is the fastest way to get something generic with your product name spelled wrong. Pinly splits it up: what has to be real, what gets built, and what AI is allowed to touch.

Layer ABuilt today

Real assets come first

Your product photos, images from your own pages, your logo, your uploads. If you sell it, people see the real thing. AI never gets to reimagine a product someone is about to buy.

product_17.png · /blog/matcha-guide hero · logo.svg

Layer BBuilt today

Built, not generated

The Pin gets assembled: your photo, the headline, your colours, your fonts, the template, the button. Laid out properly and exported at Pinterest’s 1000×1500. Same input, same output, every time.

TemplateConfig → HTML → PNG

Layer CBuilt today

Generation, only as a backdrop

Never “make a nice Pin”. Only “a warm, minimal kitchen counter, empty space on the left, no text” — and then Pinly draws your real product and real headline on top. Off unless you turn it on.

backgrounds · textures · mood · illustration

5-Minute Matcha Morning Routine

See the routine

Click a layer to take it off the Pin and put it back.

That headline is real text, not pixels. Which is why Pinly can check it fits, check you can read it, keep it in your font, and swap it for a different one without touching anything else.

One rule under all three: a Pin never shows a product that doesn’t exist, or a photo nobody took. AI backgrounds get labelled as AI in the Pin’s own record. And they stay off until you say otherwise.

The creative compilerBuilt today

Not a prompt. A brief.

Every Pin starts as a form with twelve fields. That form is why everything else works. It tells the builder what to make, and later it tells you what to change.

“make a nice Pinterest pin for my matcha powder” — one sentence. Twelve decisions made for you. None of them written down.

Creative brief
product
Organic Matcha Powder
audience
25–40, health-conscious professionals
keyword
matcha morning routine
intent
inspirational / educational
angle
5-minute morning routine
headline_pattern
headline
5-Minute Matcha Morning Routine
hero_asset
product_17.png
template
editorial_04
visual_style
minimal / natural
cta
See the routine
destination
/matcha-morning-routine

Click a pattern above to change one field and watch the rest hold still.

What that gets you

  • Something can build it. Template plus photo plus copy is a layout job, and layout jobs have right answers.
  • A test is obvious. Copy it, change headline_pattern, leave the other eleven alone. Now you know what you tested.
  • Results add up. “Which angles work for me” becomes something you look up, not something you argue about.
  • Nothing is a mystery. Six months later, every Pin can still tell you why it looks the way it does.

That’s the difference between a tool that hands you a Pin and a tool that can tell you what it did. The prompt is disposable. The record is the asset.

The quality gate

Nothing goes out unchecked

Automated posting without checks is a machine for embarrassing yourself at scale. These run on the finished Pin, not on the prompt that asked for it.

Pre-publish runBuilt today
  • Length limits

    title 62/100 · description 340/800 · alt 96/500

  • Destination URL

    https, resolves, under 2048 characters

  • Filler text

    no “lorem”, no “TODO”, no unfilled blanks

  • Text fits the template

    this headline gets cut off in editorial_04 — shorten it or pick another template

  • Still readable

    text didn’t have to shrink past the readable size

  • Contrast

    headline stands out against the background

  • Claim you can’t back up

    “clinically proven” appears nowhere on the page this Pin links to

  • Keyword stuffing

    nothing repeated often enough to read as spam

  • Image attached

    a real image, not a fallback colour block

Blocked — 1 error, 1 warning. Nothing was published.

You have to fix the error. You have to sign off on the warning yourself — a claim your site doesn’t make isn’t a formatting problem.

Also designed

Designed, not built
  • Is the product actually visible?
  • Is the logo in the right place, in your colours?
  • Does the image match what people searched for?
  • Does the headline match what the page actually delivers?
  • Is the layout too busy?

Duplicate detection

Designed, not built

Post forty near-identical Pins and you teach Pinterest that your account is noise. Comparing every new Pin against everything you’ve already published turns “is this actually different?” into a number.

cosine similarity to pin_04120.00

threshold 0.90 — above it, the variant is rejected and regenerated

Counting characters is easy. Asking whether a Pin promises something your page doesn’t deliver is the one that keeps you honest when you’re not watching.

Demand signalsBuilt today

What we measured, what we guessed, what nobody knows

Most AI keyword tools hand you a hundred phrases with confident numbers next to them. The phrases are sometimes useful. The numbers are made up — and they’re the part you’d trust most.

What you’ll never see here

12,430 searches / month

Pinterest doesn’t publish search volume through its API. If a tool shows you this number for a Pinterest keyword, it either came from somewhere else entirely, or an AI invented it.

What you get instead

High opportunityRisingStrong brand fit

Less impressive. Actually true. Each one shows where it came from and when.

Provenance on every signal
  • Historical performance

    Measured

    Your own published Pins on this topic, straight from the Pinterest analytics API.

  • Trend direction

    Measured, when available

    Pinterest has a Trends API, but it’s aimed at agencies, enterprise clients and partner platforms. So Pinly treats it as a bonus and works honestly without it.

  • Content fit

    Computed

    Do you actually have a page, product or photo that answers this?

  • Semantic relevance

    AI estimate

    An AI judging how close this term is to what you sell. Labelled as a guess everywhere it shows up.

  • Search volume

    Not available

    No source we trust, so no number. Pinly says so instead of filling the gap.

A map, not a list

Keywords only mean something next to each other. Group them and you can see where you’re already covered, where you sell something but have written nothing, and where you haven’t touched a whole branch.

matcha
iced matcha
strawberry matcha
matcha latte
matcha energy
matcha focus
matcha vs coffee
morning routine
healthy habits
wellness routine

Opportunity score

Relevance×Demand signal×Content fit×Historical performance×Commercial intent

Every part of that shows where it came from. Multiply guesses and you still have a guess — so the score ranks your keywords against each other. It never pretends to be a real quantity.

The honest version screenshots worse and works better. You can act on “this branch is growing and you already sell three things that fit it”. You can’t act on a number somebody made up.

Pin DNA

A Pin isn’t an image. It’s a record with an image attached.

“pin_123 · CTR 4.7%” tells you something worked. It doesn’t tell you what. Save the 44 things below and every Pin you post teaches the whole account something. 30 are saved today. The dashed ones are planned.

5-Minute Matcha Morning Routine

pin_0431

published · Tuesday 09:00 · Healthy Morning Ideas

Content

  • topic
  • keyword_cluster
  • source_page
  • product
  • search_intent
  • content_angle
  • content_pillar
  • funnel_stage

Copy

  • headline
  • headline_pattern
  • headline_length
  • cta
  • description
  • number_present
  • question_present
  • benefit_present
  • sentiment
  • reading_level

Visual

  • template
  • image_type
  • background_type
  • visual_style
  • brand_palette
  • typeface
  • contrast_ratio
  • image_embedding
  • dominant_colors
  • text_density
  • product_size
  • logo_position

Distribution

  • board
  • weekday
  • hour_window
  • campaign
  • variant_group
  • season

Outcome

  • impressions
  • saves
  • pin_clicks
  • outbound_clicks
  • engagements
  • ctr
  • save_rate
  • conversions
recorded todaydesigned, not yet written

Notice the results column is the shortest one. Results are easy — Pinterest gives them to you. The hard part is having written down, months earlier and for every single Pin, exactly what you did differently.

Attribution

A winning Pin doesn’t explain itself

This is why “AI learns from your best posts” is mostly worthless. Two Pins did differently. Eight things about them were different. Picking one and calling it the reason is a guess in a nice outfit.

Uncontrolled comparison

Pin A · CTR 5.2%  vs  Pin B · CTR 1.8%

backgroundredwhite
headline_patternlisticleeducational
image_typeproductlifestyle
productmatcha tinstarter kit
keyword_clustermatcha energymatcha recipes
boardHealthy DrinksMorning Ideas
weekdayTuesdaySaturday
hour_window09:0020:00

Eight differences. One result. You’ve learned nothing.

“Red works better” is what a lazy tool tells you here. It could just as easily be the topic, the board, the product, or the time of day.

Controlled experiment

one variable open, seven held constant

Built today

One difference. The result means something.

And it is still just a hunch until a second run repeats it. If it does not repeat, it was noise.

This is why the brief has named fields and why every Pin carries dozens of them. Not for neatness. You can only hold seven things steady if you know what your seven things are.

The learning engine

Simple math first. Fancy math when it earns it.

Nobody should start with a neural network. With forty Pins of history, an average with a sanity check beats anything trained on the same forty Pins — and it’s honest about what it doesn’t know. The clever stuff arrives when there’s enough data to earn it.

  1. Phase 1Built today

    Descriptive statistics

    Compare each group against your own average, and check there’s enough data for that to mean anything. Boring. Useful in week one.

  2. Phase 2Designed, not built

    Bayesian estimates

    Handle small samples properly, so a promising group with nine Pins is neither thrown away nor over-believed.

  3. Phase 3Designed, not built

    Feature importance

    Rank which of your choices actually drive the result, instead of looking at one thing at a time.

  4. Phase 4Designed, not built

    Contextual bandits

    Push the next batch towards what is winning, while still testing the things you’re unsure about.

Eight to exploit, two to explore

Designed, not built

Copy your winner forever and you stop finding new ones. Then Pinterest moves and you’re stuck. So part of every batch goes to ideas the data doesn’t back yet.

8 exploit — what the evidence favours2 explore

Most of the batch on what works, a slice kept back for what might.

Three sources, shifting weight

Designed, not built

Day one you have no history, so Pinly leans on what generally works in your industry. Every Pin you post shifts the weight towards your own data. The longer you use it, the more it’s about you.

day one0 months of your own data
  • What works everywheregeneral Pinterest patterns70%
  • What works in your industryfood, beauty, interiors25%
  • What works for youmeasured on your own Pins5%

The AI explains the number. It never makes it up.

Built today

// computed in the database

segment: headline_pattern = listicle

lift_vs_baseline: +38%

n_pins: 47

n_impressions: 12,410

confidence: high

“Pins with numbered listicle headlines earned a 38% higher click-through rate than comparable educational headlines over the last 60 days, most strongly in the Healthy Drinks cluster. Next test: three more listicle headlines on the same template and image style.”

47 Pins · 12,410 impressions · high confidence

The AI gets a finished result and does two things: writes it in plain English, and suggests what to test next. It never gets asked what the number is. That’s the only reliable way to stop it inventing one.

Every claim on your dashboard can be traced backwards: which query, how many Pins, how sure. If Pinly can’t show you that chain, it doesn’t make the claim.

The flywheel

All of it, as one loop

A generator gives you Pins. A scheduler gives you a calendar. Neither gets better at your account. Everything above exists to close this circle — the last step feeds the first.

1Understand2Spot3Create4Publish5Measure6Learn
  1. 1

    Understand. Read your site. Work out what you actually sell.

  2. 2

    Spot. Turn that into real Pin ideas, each with a reason attached.

  3. 3

    Create. Write the copy. Design the Pin. Make versions that differ.

  4. 4

    Publish. Schedule it. Post it through the official Pinterest API.

  5. 5

    Measure. Pull the numbers daily and tie them to the right groups.

  6. 6

    Learn. Find what beat your average. Feed it back into step three.

Step six is the one everyone skips. Skip it and you guess forever — because a Pin that never recorded its angle, headline style, template or board can’t tell you anything later.

What you get back

Pinterest tells you how a Pin did. Pinly tells you which of your choices did it — because every Pin remembers the angle, headline style, template, topic and board it came from.

Example workspace · last 90 daysIllustrative figures

Impressions

48,210

Outbound clicks

1,642

Click-through rate

3.4%

Saves

2,981

Impressions per week — the stacking effect from the chart above, on one account

Click-through rate by headline style, against this account’s average

How-to

14 Pins · 4,120 impr.

4.1%+38%

Listicle

11 Pins · 3,480 impr.

3.2%+6%

Question

9 Pins · 2,900 impr.

2.1%−31%

Quote

2 Pins · 210 impr.

not enough data

How-to headlines beat your average by 38%

Based on 14 Pins and 4,120 impressions over 90 days. Suggested next step: four versions on your three best topics, using the how-to style.

Look at the fourth row. Two Pins and 210 impressions isn’t a finding, so Pinly won’t call it one. No green arrow. No confident percentage. Every screen in the app works this way, and every number shows where it came from and when.

The unglamorous parts, done properly

Anyone can put a text box next to a model. The work is in everything around it.

Ten Pin templates, on brand

Editorial, listicle, recipe, comparison and more. Built at Pinterest’s 1000×1500 in your colours, your fonts, your logo.

Variants that actually differ

Each version takes a different angle and a different headline style — not the same sentence rearranged. So the test actually tells you something.

Scheduling that survives you

A queue that retries, recovers, and shows you what failed. Double-click all you want. It can’t post twice.

Quality gates before publish

Cut-off headlines, weak contrast, filler text and claims your site doesn’t back up. All caught before anything goes out.

What it won’t do

Most of the work in a tool like this is deciding what not to claim. Three rules, everywhere:

No invented numbers

Every score shows where it came from. Pinterest data says Pinterest. An AI guess says guess. There’s no third option.

No pattern without evidence

Under three Pins or five hundred impressions, you see “not enough data”. Pinly would rather say nothing than call noise a trend.

You approve what ships

AI suggests, you decide. That’s the default. Autopilot exists, stays off until you turn it on, and stops the second you turn it back.

Questions

Do I need a Pinterest account to start?+

No. Analyse your site, generate ideas and design Pins right away. You only need Pinterest when you want to publish and pull results.

Does it post without asking me?+

Only if you turn on autopilot, which is off by default and capped by limits you set. Otherwise nothing leaves your account without you.

Where do the Pin images come from?+

Templates built in your brand colours, plus images from your own site or your uploads. AI images are optional, off by default, and never used to fake a photo of a real product.

How does it decide what works?+

Every Pin saves its angle, headline style, template, topic and board. Results get split by those, with a floor so small samples don’t get treated as findings.

Some sections say “designed, not built”. Why show them at all?+

Because the plan is the pitch, and hiding the unfinished parts would be exactly the thing this product exists to avoid. “Built today” means it runs right now. “Designed, not built” means it’s specced and scheduled. Nothing here is described as done when it isn’t.

Why are there no numbers about Pinterest on this page?+

Because we’d be quoting them from memory. Pinly won’t show you a number it can’t source, and a landing page that broke that rule would be a bad advert for it. Everything here is about how the platform works, which you can check yourself.

Is this affiliated with Pinterest?+

No. Pinly posts through the official Pinterest API and follows their developer policies, but it’s an independent product.

Pin something worth saving.

Start with your website. Pinly does the reading, the writing and the measuring. You keep the final say.

Start creating