{
"title": "I Ran a 30-Day GEO Experiment on Our SaaS Blog — Here's What Actually Happened",
"content": "Last Tuesday I pulled our analytics and noticed something weird: our organic traffic from Google was flat, but mentions of our product in ChatGPT and Perplexity responses had tripled over the past two months. No campaign. No PR push. Just... showing up.\n\nSo I decided to run a controlled experiment. For 30 days, I rewrote 10 of our highest-traffic blog posts using GEO principles — not SEO principles, GEO principles — and tracked whether AI engines started citing us.\n\nHere's exactly what I did and what changed.\n\n## The Setup\n\nI picked 10 posts that already ranked well on Google (top 3 positions) but had zero presence in AI-generated answers. These were how-to guides and comparison posts in the project management space — exactly the kind of content people ask AI about.\n\nThe control group was the original versions. The test group got GEO-optimized versions. Same topics, same word count range, same factual accuracy. Different structure.\n\n## What I Actually Changed\n\n1. I moved the answer to the top.\n\nEvery original post had a standard intro: hook → background → thesis → then the answer. I flipped it. The test versions led with a direct, 2-3 sentence answer to the core question, then expanded.\n\nFor example, our post \"Best Project Management Tools for Remote Teams\" used to open with a paragraph about the rise of remote work. The new version opened with:\n\n> For remote teams under 20 people, the best project management tool is Linear for async-first workflows and Notion for teams that need documentation baked in. For teams over 50, it's Jira or ClickUp depending on whether you need custom workflows.\n\nThat's it. No throat-clearing.\n\n2. I added explicit \"according to\" attribution.\n\nAI engines love sourced claims. I went through every statistic and claim and added specific attribution. Not \"studies show\" — \"According to the 2024 State of Remote Work report by Buffer, 67% of remote teams cite communication as their top challenge.\"\n\nI added 3-4 attributed claims per post. This alone felt like the biggest lever.\n\n3. I restructured for snippet extraction.\n\nI broke every post into clearly labeled sections with descriptive headers. I added comparison tables with specific criteria. I included \"When to use X\" and \"When to use Y\" decision frameworks.\n\nThe goal: make it trivially easy for an AI model to extract a coherent, useful chunk without needing to paraphrase heavily.\n\n4. I added a \"Bottom Line\" section at the end.\n\nNot a summary — a recommendation. \"If you're a startup under 10 people, use Linear. If you're an enterprise team with compliance needs, use Jira. Here's why.\"\n\n## The Results After 30 Days\n\nOf the 10 GEO-optimized posts, 7 now appear as cited sources in AI-generated answers for their target queries. I tested this by running the original search queries through ChatGPT, Perplexity, and Google's AI Overview.\n\nThe 3 that didn't pick up? Two were on highly competitive keywords where established domains (Asana's own blog, Atlassian) dominate. One was a post where I was too conservative with the rewrite — I kept too much of the old structure.\n\nHere's the traffic breakdown:\n\n- AI-referred visits: 340 over 30 days (tracked via UTM parameters and referral data)\n- Featured snippet appearances: 4 of 10 posts now appear in Google's AI Overview\n- Direct citation rate: When I asked Perplexity \"what are the best project management tools for remote teams,\" our post was cited in 3 out of 5 test runs\n\nFor context, these same 10 posts were getting about 2,100 organic visits/month from Google before the experiment. The AI-referred traffic is a 16% bump — not life-changing, but it's a new channel that didn't exist for us two months ago.\n\n## What I Got Wrong\n\nI initially thought GEO was just \"SEO but shorter paragraphs.\" It's not. The content needs to be structured for a fundamentally different consumption pattern. AI engines don't scroll. They extract.\n\nI also underestimated how much attribution matters. My first two rewrites had the same structure but vague sourcing. They didn't get picked up. Once I added specific \"according to\" citations with named sources, the pickup rate jumped.\n\nIf you're trying to figure out where to start, I'd recommend running a GEO Audit Tool on your top posts before rewriting anything. It'll show you which of your pages AI engines are already reading and which are invisible.\n\n## The Bigger Picture\n\nHere's what's keeping me up at night: the traffic from AI engines converts at about half the rate of organic search traffic. The visitors are earlier in the funnel — they're asking questions, not looking for solutions yet.\n\nBut the volume is growing fast. Our AI-referred traffic in month one was 89 visits. Month two was 210. Month three was 340. That's a compound monthly growth rate of about 56%.\n\nIf you want to understand how this channel compares to traditional organic, I wrote a breakdown of GEO vs SEO that maps out where they overlap and where they diverge.\n\n## What I'm Doing Next\n\nI'm expanding the experiment to 30 more posts. I'm also testing whether adding structured data (FAQ schema, HowTo schema) increases citation rate. Early signal: yes, but only marginally. The content structure matters more than the markup.\n\nOne thing I'm NOT doing: I'm not rewriting for AI at the expense of human readers. The test posts still need to work for someone who lands on the page directly. The trick is that GEO-friendly structure — direct answers up front, clear attribution, decision frameworks — actually makes the content better for humans too.\n\nIf you're sitting on a library of SEO-optimized content that's invisible to AI engines, you're leaving traffic on the table. The gap between \"ranks on Google\" and \"cited by ChatGPT\" is smaller than you think. You just have to structure for extraction, not just ranking.\n\nI'm tracking everything in a public spreadsheet if you want to see the raw data. DM me and I'll share the link.",
"tags": ["GEO", "AI search", "content strategy", "SaaS marketing", "ChatGPT optimization"],
"summary": "Ran a 30-day GEO experiment on 10 SaaS blog posts: 7/10 now cited by AI engines. Here's the exact framework and results."
}