Breaking IconScout’s organic ceiling - then building past it

Context

IconScout is LottieFiles’ creative marketplace - icons, illustrations, 3D, Unicons, Lottie - used by 6.8 million people worldwide. The product is the catalog. Millions of product listing pages.

I managed SEO here while also owning LottieFiles. Same company family. Different graph. Same lesson: marketplace SEO is an architecture problem wearing a content costume.

By early–mid 2024, IconScout had hit a wall most large catalogs know well. Clicks per 28 days stalled under ~500k. Demand existed. Indexation and discoverability did not keep up.

Three bets broke that ceiling - and opened a second surface beyond the library.

Problem

PLPs were islands. Millions of listing pages, thin connections. Crawlers burned budget. Long-tail relevance leaked. Related assets and trending queries were not systematically surfaced.

Hero copy was generic where competitors were specific. Competitor research showed high-performing pages with custom, keyword-tailored hero descriptions. Ours often weren’t. At this scale, hand-writing every page is fantasy - unless you prove a pattern, then automate it.

Tools and AI features had product energy and almost no SEO surface. IconScout was expanding past “asset library.” Search still treated it like one.

The constraint wasn’t more blog posts. It was: connect the catalog, write the money templates at scale, and give new product bets pages that can rank.

Approach

1. PLP internal link architecture

We rebuilt linking from search behavior - not editorial taste.

Internal search data drove dynamic trending search links on PLPs. Google Search Console relationships powered people also search for sections - real query adjacency between product pages.

The goal: crawl paths and contextual relevance that scale with millions of URLs. Architecture, not a spreadsheet of one-off links.

Indexation sat next to clicks as a first-class metric. In marketplace SEO, indexation often leads traffic by months. Celebrate clicks alone and you kill the program early.

Engineering owned the infrastructure. Content and design made the modules legible. I owned prioritization: which templates, which clusters, which thresholds before expanding.

2. PLP hero content - prove, then scale

Competitor pages winning our keywords had custom heroes. So we ran it in phases.

Phase 1: update heroes on pages already competing for those keywords. Traffic and rankings moved.

Phase 2: expand into untapped categories. Same pattern held.

Phase 3: I trained a custom GPT that takes a page URL and returns on-brand hero copy - so thousands of PLPs could refresh without a forever editorial queue.

Content lifts stick when the graph can support them. Linking opened the ceiling. Heroes filled the pages that could finally earn the click.

3. Tool and AI landing pages

New tools and AI features needed high-intent landing pages - not buried product notes.

I led a cross-team template: content quality, technical SEO, UX, aligned to brand and product marketing. Built to reuse as the suite grew.

The bet wasn’t a one-off launch page. It was a system that positions IconScout as a creative platform in search - not only a library of assets.

Outcome

~2× organic clicks from the linking program - past the old 500k / 28-day ceiling, through 700k and 800k milestones in October 2024, and into the ~900k range (886k toward a 900k Google Search Impact window at peak tracking).

~80% click increase on PLPs touched by the hero content program - the scaled Phase 3 model after the phased proof.

0 → 20k clicks on tool landing pages - a new organic surface for tools and AI, with a template ready for what ships next.

For leadership, the story shifted from “marketplace SEO is tapped out” to “discoverability and money-page systems were the bottleneck.” That reframing unlocks engineering time.

What I’d do differently

I would publish an inheritance playbook earlier for how new asset types pick up linking rules. Marketplaces ship verticals constantly; without it, every launch reopens architecture debates.

I would pair the linking rollout with a cannibalization monitor from day one. When related PLPs gain crawl equity, near-duplicate clusters fight. We managed it - a dashboard would have cut the fire drills.

And I would have wired tool-page launches to the same release clock as product. When a feature ships, the SEO template should ship with it - not in a later pass.