Much of the AI-generated content that flooded the web in 2024 and 2025 has now disappeared from search results. It ranked for a few weeks, Google’s helpful content systems caught up, and the rankings collapsed. Most of that spending produced content debt instead of organic traffic.
Programmatic SEO drew an outsized share of those losses. SEO practitioner Mark Williams-Cook, who runs AlsoAsked and the agency Candour, has been vocal about the pattern Google’s update targeted. “I saw lots of sites doing what I’ll loosely call ‘programmatic SEO’ with People Also Ask data — they’d take any topic, scrape 50 PAA questions, and then, in a basic Q&A format, list the answers and generate tens of thousands of pages,” Williams-Cook wrote. “If a site has lots of pages that fit that pattern, it’s probably a signal that it’s not a great site.” That pattern matched what Google’s spam team formalized in its scaled content abuse policy.
SEO consultant Aleyda Solis frames the alternative directly. “Although it’s possible to leverage AI for content, a personalized editorial and optimization workflow is required to ensure quality, originality, and expertise by integrating unique brand insights and first-party data, which is exactly what AI platforms are likely to cite,” Solis told Search Engine Journal. The question for any programmatic build is whether it looks more like the pattern Williams-Cook described or the one Solis points to. Alpaca Health is one example of the second.
The Series A healthcare startup, founded in 2024 and backed by South Park Commons, Adverb Ventures, and Tau Ventures, connects families seeking autism therapy to the Board Certified Behavior Analysts who deliver it. The company hired Hassan Rashid, now managing editor at GrowthX AI, to rebuild a programmatic content footprint that had stopped ranking. Alpaca Health had been running a generic vendor template for its city and state pages, the kind that swaps city names and a few local fields around a shared shell. Rashid rebuilt the entire programmatic layer from scratch and shipped 238 location pages in a thirty-day window. The numbers since have been strong.
Within just over a month of publishing, the city and state pages Rashid rebuilt were driving almost 60 percent of all non-homepage organic traffic to Alpaca Health’s family intake form. Over a four-week window, those pages generated 114 intake form clicks. The Texas state page became the second-highest-converting non-homepage URL on the entire site. Across the same window, the city and state layer drew 395 organic clicks and more than 43,000 impressions from Google, with the Texas state page alone accounting for 23 clicks and 2,114 impressions. Rankings climbed alongside the traffic and conversions, with multiple pages now ranking on the first page of Google.
After the rebuild closed, Alpaca Health’s CEO adopted the system Rashid built as the company’s official content strategy for the next five state launches: North Carolina, Hawaii, Nebraska, Vermont, and Nevada.
Rashid’s system runs on Cursor and Claude Code. Each engagement gets a dedicated artifact set the workflow reads on every draft: writing guidelines, a company context document, a proofreader checklist, brand-voice references. Through Anthropic’s Model Context Protocol, the workflow also connects to Semrush, Google Search Console, and the live SERPs for the queries that Alpaca Health is targeting.
The programmatic layer is what puts Alpaca Health’s rebuild on the primary-data side of that line. For Alpaca Health’s city and state pages, the workflow pulls real-world data per page: US Census demographics, local school district records, state-specific insurance citations, and the names of local crisis lines and community resources. Each was chosen for what parents navigating autism care actually need to know: which insurance plans cover therapy in their state, which school districts support neurodivergent students, and where to call when they need help between appointments. Every page then routes through a human-in-the-loop editorial review and ships with structured author and reviewer schema for Google’s E-E-A-T signals. The output is city-level content built from primary data rather than rephrased competitor pages, which is the line Williams-Cook drew when describing what got penalized.
Those conditions trace back to March 2024, when Google folded its helpful-content signals into its core ranking system over a 45-day rollout and later estimated the update cut 45 percent of low-quality, unoriginal content from results. AI authorship correlated with the failures, but the underlying problem was the missing editorial layer.
The Alpaca Health outcome is not isolated. The other engagements Rashid produces content for run on the same infrastructure, applied to editorial articles rather than programmatic pages. The workflow looks different across the two, but the human-in-the-loop review and E-E-A-T principles that lifted Alpaca Health’s programmatic results carry directly into the editorial work, with demonstrable expertise and named author authority on every piece.
At a venture-capital firm, each article was built around what the firm had learned from years of conversations with founders rather than repackaged industry takes that any competitor could copy, and the workflow routed every draft through partners and operators who had actually been in those rooms. Across the program’s first quarter, the firm’s content went from effectively unranked to compounding organic traffic: monthly clicks grew 33x, from 34 to 1,108, and monthly impressions grew 14x, from 57,000 to 820,000, while sessions reached 1,672 a month and 2,777 cumulative across the three months, 4.8 times the content program’s optimistic forecast. The same work carried into AI search, where the firm rose from a negligible share to the second-highest AI-assistant visibility in its competitive set, and referral traffic from LLMs grew 183 percent month over month.
The same pattern repeats at an enterprise B2B SaaS engagement, where every article was anchored in product-specific knowledge the company’s engineers and PMs had earned firsthand, with named experts reviewing each piece before it shipped. Across 44 published articles, the content now draws 1,731 organic clicks, more than 660,000 impressions, and 5,751 organic sessions a month, with monthly clicks more than doubling and impressions up 174 percent over the prior month. AI assistants have become a channel of their own, sending 174 referred sessions a month across 26 of the articles, up from 23 a month earlier, with the work earning 794 of the firm’s tracked LLM citations, more than a third of the total. In the same window, the content drove 29 commercial conversion events, spanning free-trial requests, demo completions, a pricing inquiry, and lead-generation form submissions.
“Most AI content collapses for the same reason, whether it’s a thousand programmatic pages or one in-depth article: the model is just regenerating what competitors already published and Google already indexed,” Rashid said. “The fix doesn’t change with the format. Once every page carries primary data, local records on a city page or firsthand expertise in an article, and a human is in the loop actually reviewing the work, you give Google something it wants to surface.”
Three design choices carry most of the weight in Rashid’s system. The first is selecting data sources where each page has something only that page can say, like local insurance citations or specific school district records, rather than rephrasing what competitors already publish. The second is building the human-in-the-loop review around the specific failure patterns AI drafts produce, not generic copy-editing. The third is connecting the SEO research, the SERP data, and the writing prompt to the same context, so the workflow does not run on conflicting assumptions.
The premise underneath the workflow is simple: Google ranks what readers actually find useful, so the work is engineering pages around real questions and real expertise rather than gaming a signal. Rashid built the system from a product management background. He spent two years as an associate product manager at Addepar (managing more than $9 trillion in client assets), the wealth management platform Joe Lonsdale co-founded in 2009. He joined GrowthX AI in September 2025, where he works on content production, AI workflows, and content strategy for companies including Ramp and Vercel. The startup raised $12 million in its Series A last year.
The Alpaca Health rebuild illustrates a structural shift inside venture-backed B2B content marketing. The work that keeps marketing teams ranking comes down to editorial judgment, real expertise, and content that genuinely helps the reader. AI can scale that work, as the Alpaca Health programmatic results show, but only when paired with human-in-the-loop review and the author-authority signals Google’s E-E-A-T model is built to reward. Rashid’s work on Alpaca Health is one example of what that pairing looks like in practice.
