TextWordCount

TextWordCount exists because the existing word counters were bad. Not broken — just frustratingly slow, ad-heavy, and trust-hostile. They paste your text into a server somewhere and return a number. For anyone working with sensitive content — legal drafts, unpublished manuscripts, client work — that's an unacceptable tradeoff.

The brief was simple: build the word counter I actually wanted to use. Browser-only processing (technically true, not marketing copy), every useful metric on a single page, and genuinely free — no paywall, no premium tier, no credit system.

Privacy-first architecture as the primary differentiator. The privacy claim is built into the architecture, not the marketing copy — every analysis function runs in JavaScript in the browser tab. This is documented in the /methodology page and verifiable by anyone who inspects the network traffic.

Depth over page count. An earlier version of the site generated close to 900 template pages across tool × use-case, tool × language and tool × city combinations. They were retired in 2026 and folded back into their parent tool pages: a smaller set of hand-written tools and articles that each earn their place beats a large surface of near-duplicate pages.

The multi-tool suite launched with word count, character count, reading time, keyword density, sentence counter, paragraph counter, case converter, text summarizer, grammar checker, and paraphrase tool — all in a single unified interface. The design decision to put everything in one place, rather than separate tools on separate pages, reduces friction for the core user journey: someone who needs “word count plus reading time plus keyword density” gets all three without navigating.

The Gumroad storefront integration at /store-app is the revenue layer. Digital products — Claude Code skills, SEO toolkits, AI agent kits — are sold through a 32-language storefront that matches the site's international audience. The storefront is programmatically translated, so a Turkish visitor sees Turkish copy; a Portuguese visitor sees Portuguese copy.

The programmatic layer is the part that did not survive. Roughly 900 generated pages covered tool × use-case, tool × language and tool × city combinations, and on paper they read like cheap coverage of intent-driven and international queries. In practice most of them said the same thing with a different noun swapped in, and over half the language pages fell back to English body copy. They were removed in 2026 and 301'd into their parent tool pages. The lesson worth keeping: page count is not a strategy, and a template that cannot say something genuinely different per variant should not be shipped per variant.

Schema markup on every page is optimized specifically for AI Overview citation. The FAQPage schema answers questions in the exact format that LLMs prefer to quote. The HowTo schema provides step-by-step instructions that AI systems can cite verbatim. TechArticle schema identifies the page as a technical reference rather than a blog post — which signals authority to both Google and AI crawlers.