Building a Content Cluster and FAQ Schema to Rank an Indie Product on Bing

I shipped NorthLab Folders — a browser extension that adds a real folder sidebar inside claude.ai and chatgpt.com — and within days of submitting it to Bing Webmaster Tools, I started seeing impressions on long-tail Claude-related queries. Not rankings, not clicks — impressions. But that signal was enough to tell me something was working, and I wanted to understand why and lean into it deliberately.

What followed was a few weeks of reading keyword data, writing focused guides, wiring up structured data, and thinking carefully about what makes content earn trust from search engines that are increasingly powered by the same kind of language models my product targets.

Here's what I did and what I'd do again.

Reading the Keyword Data First

I didn't start with a content plan. I started with data.

Bing Webmaster Tools and Google Search Console both expose the queries that triggered your pages. I went through both carefully, not looking for high-volume head terms (there weren't many), but for query clusters — groups of phrases that shared a clear distinct intent even when they used different words.

What I found: people weren't just searching "Claude folders." They were searching things like "how to make a folder of chats in Claude," "can Claude read a folder," "Claude folders on Mac," "how do Claude Projects work." Each of those is a different question with a different answer. Lumping them onto one page would mean serving none of them well.

That's when I committed to a hub-and-spoke structure instead of one long page trying to do everything.

The Hub-and-Spoke Architecture

The pillar page — How to Add Folders to Claude — covers the general case. It exists to answer the broadest version of the intent and internally link to everything else. It's not the longest page; it's the most authoritative one, because it connects the cluster.

Each spoke targets a single intent specifically. How to make a folder of chats in Claude is one example — it answers exactly that question, honestly (Claude has no native folder primitive; what exists is Projects, which is different and has real limitations), and then explains what the extension actually does differently.

A few principles I held to:

• One primary intent per spoke. The temptation is to pad each guide with tangential content. I resisted. Focused pages rank for focused queries.
• Honest about limitations. Claude's Projects feature is not folders in the way most people mean. It doesn't let you pin individual chats. Free plans are capped at five projects. There's no nesting. I say this clearly. Searchers asking these questions deserve an accurate answer, not a bait-and-switch.
• Internal links between every piece. The pillar links to every spoke. Spokes link back to the pillar and to related spokes where the connection is natural. This matters for crawl authority, but more practically it matters because a reader who landed on the wrong guide should be one click from the right one.

FAQPage JSON-LD: A Bing-First Play

I added visible FAQ sections to the key guides and backed them with FAQPage JSON-LD structured data. Every FAQ is rendered on the page — the schema mirrors visible content, never hides it.

Here's the honest nuance: Google deprecated FAQ rich results for most sites in late 2023. They still process the schema and it may factor into AI Overviews, but you're not getting the expandable FAQ boxes in standard Google SERPs anymore (except for authoritative health and government sites). I knew this going in. The play was explicitly Bing-first — Bing Webmaster Guidelines still support FAQ rich results, and Bing Copilot uses structured data as a signal when surfacing answers.

So the FAQ schema was a deliberate Bing investment, not a Google one. I didn't treat it as a magic ranking lever; I treated it as a way to give Bing's crawler a clean, machine-readable summary of what each guide answers. The structured data reflects real questions that showed up in my Webmaster Tools queries — I wasn't guessing at what to ask.

Writing for Language Models (Without Keyword-Stuffing)

The thing that changed how I think about SEO content: Google's own guidance for AI-optimized content explicitly says not to keyword-stuff, because language models understand synonyms and semantic relationships. You don't need to repeat "Claude folders" seventeen times for a model to understand the page is about organizing Claude conversations.

What models reward instead is first-hand, specific, non-commodity content. Observations from actually building and using the thing. Honest acknowledgments of limitations. Concrete details (NorthLab Folders is local-first; it stores folder names and chat IDs in extension storage, not on any server; there's no cross-device sync because there's no backend — that's a real tradeoff worth explaining). That kind of content is harder to replicate by scraping and remixing existing pages, which is exactly why it has a better shot at earning a citation in an AI-generated answer.

I applied this to every guide in the cluster. The goal was to write things that only someone who built or deeply used the product would know to say.

What's Actually Happening

I'm going to be straight about results: impressions are climbing on Bing for the long-tail queries I targeted. Rankings are still maturing — some guides are in the top ten for their primary term, others are fluctuating. I haven't measured enough time to draw strong conclusions about the cluster's aggregate effect versus individual page quality.

What I can say with confidence: the queries I'm getting impressions for are the exact queries I wrote spokes for. That's the strategy working as intended, even if the conversion story is still early.

Google is a slower story. The domain is relatively new, trust builds over time, and FAQ rich results are off the table. I'm writing the guides to earn AI Overview citations rather than traditional blue-link positions, which is a longer game.

What I'd Tell Another Indie Dev

If you're building something in a niche where the core product category is actively being searched — even in small volumes — the content cluster approach is worth the investment. One pillar page plus four or five focused spokes beats one long-form "ultimate guide" almost every time, because search intents don't cluster the way we wish they did.

Add FAQ schema if Bing is a realistic traffic channel for your audience. For Claude-adjacent products, Bing Copilot is increasingly how people find and evaluate tools — it's worth treating Bing as a first-class platform, not an afterthought.

And write the content yourself, from your own experience building the product. Not because it feels more authentic (though it does), but because that's what the models running search are actually optimizing for now. Commodity content is free to generate; first-hand observations aren't. That gap is the only durable SEO advantage an indie maker actually has.