The Growth Newsletter #325How to engineer your AI search flywheel Welcome to the third and final edition of our AI search series. In Part 1, we made the case that AI search is the first truly disruptive distribution platform we’ve seen in well over a decade. In Part 2, we laid out a step-by-step playbook for building your foundation. Now, in Part 3, we’re talking about what happens after the foundation is in place: how to engineer an AI search flywheel. Many of our clients are moving past the foundation phase. Citations are live. Visibility is established. We’re entering into the flywheel phase. New citations showing up for queries nobody targeted. Brand search volume climbing without paid spend behind it. Third-party sites referencing these companies unprompted. This is the phase where AI search stops being a shiny new channel and starts showing why we believe it might be the single biggest distribution opportunity for startups. Here’s how we’re engineering AI search flywheels for our clients over at Saturation.
Brought to you by HighLevel HighLevel's all-in-one platform connects marketing, sales and automation into one command center. Every workflow gets faster when AI handles the busywork. From lead capture to follow-up, AI helps keep messages personal and perfectly timed. No code, no chaos… just results. Whether you manage one brand or a hundred, HighLevel automates the tasks that used to require entire teams. Now, you can build, grow and lead with the only AI system built for businesses like yours. Start your 30-day free trial here.
Brought to you by Omnisend Email & SMS marketing so good, it's boring Let's be real. Email marketing is not the "sexiest" member of the marketing family. And that's okay. As a business owner, you have enough "exciting" stuff happening. Let Omnisend be the reliable, stable and effective platform you don't need to worry about. Check out Omnisend and take a nap Why AI search plays by different rulesBy Zach Boyette, Managing Partner @ Saturation. In SEO, your brand helps you win individual keyword battles faster. A strong domain ranks new pages more easily. But each keyword is still its own fight. You need a dedicated page, on-page optimization, and backlinks for each term you want to rank for. AI search works differently. In AI, your brand IS the battle. LLMs build associations between brands and categories during training. When someone asks a question, the model activates those associations and selects which brands belong in the answer before it goes looking for URLs to cite. Brand selection first, citation retrieval second. The data backs this up. Seer Interactive found that when a brand is mentioned in an AI response, its content gets cited 53.1% of the time. When the brand isn’t mentioned, that drops to 10.6%. Kevin Indig’s analysis of 1.2M ChatGPT citations found that 32.9% of cited pages appeared only in follow-up queries the content was never optimized for. AI expands your citation surface beyond what you targeted. This is where the compounding kicks in. Every citation, third-party mention, and brand reference strengthens the model’s association between your brand and your category. That association gets baked in on the next training cycle. Rand Fishkin’s research found that top brands in any category appear in 55-77% of AI responses regardless of how the prompt is phrased. Once you’re in the model’s consideration set for your category, you stay there across the full query space. That doesn’t happen in SEO. You don’t rank for keywords you didn’t build a page for. So what does the flywheel actually look like in motion? When AI cites you, users start Googling your brand, creating new search signals. Third-party sites notice you’re being recommended and mention you in their own content. And you get data back: which queries triggered the citation, how the model described you, what competitors appeared next to you. Each of those outputs feeds the next cycle and strengthens the brand-category association that got you cited in the first place. The rest of this piece is about how to build and sustain that system. Build the content architectureIndividual pages get individual citations. But clusters (groups of interconnected pages covering a topic from multiple angles) train AI to associate your brand with an entire category. Comparative listicles alone account for 32.5% of all AI citations, according to cross-industry citation analysis. That's a third of all citations going to one format. Comparison pages are the highest-leverage move you can make. Someone asking "X vs Y" is actively making a decision. ClickUp built dedicated pages comparing themselves against Asana, Monday, Notion, and Jira, a strategy that multiple growth analyses have credited as a core driver of their early traction while bootstrapped. You don’t need ClickUp’s resources. Three pages, your brand vs. your two closest competitors, creates a cluster AI can reference every time someone asks about your category. Structure matters: lead with a clear verdict, use comparison tables, and include specific statements about who each tool is best for. Not feature checklists. Buying guidance. Use case pages are your second priority. Deel publishes dedicated pages for each use case: global hiring, payroll, equipment shipping. A single product page gets cited for one generic query. Ten use case pages mean you appear for ten different buyer questions, each closer to a purchase decision. For startups, this is where you outperform: three deeply specific use case pages beat one generic product page from an enterprise competitor. Industry and topical authority pages are slower burns. They build brand association rather than driving direct conversions. Maze publishes UX research guides covering strategy, methods, and tools. These don’t close deals, but they teach the model that Maze and UX research are synonymous. Worth building once you’ve covered comparison and use case clusters first. Not before. The operating cadenceA flywheel only works if it keeps spinning. Here’s the cadence we run at Saturation, modified for teams that don’t have a dedicated AI search person. Monthly reconnaissance (2-3 hours). Run your top 15-20 buyer queries through ChatGPT, Perplexity, Gemini, and Claude. Manually. Yes, in 2026, the most sophisticated AI search strategy still starts with copy-paste into four browser tabs. We've tried to make this sound more impressive in client presentations. It isn't. But it works. Track three things in a spreadsheet: (1) whether you appear, (2) how you're described, and (3) who else shows up next to you. Compare to last month. If ChatGPT describes you as "a good option for small teams" but you're targeting enterprise, your content is positioning you wrong. That's not a content problem. That's a brand perception problem, and now you can actually see it happening in real time. Tools are emerging that automate this. Otterly AI, Peec AI, and SE Ranking Visible all track AI citations across platforms with daily refreshes. Worth the spend once you’re past the first 2-3 manual cycles and know which queries actually matter. Biweekly content (1-2 pages). Based on what your recon shows, publish one or two new pages every two weeks targeting specific queries where you’re absent or weak. Comparison and use case pages first. Not blog posts. Structured answer pages. The temptation is to try covering everything at once, but two focused pages every two weeks will outperform eight mediocre ones per month every time. Ongoing distribution. Get your positioning reinforced in directories, review sites, Reddit, and any third-party source AI already cites for your category. Multi-platform presence across 4+ channels makes content 2.8x more likely to appear in ChatGPT. AI doesn't just read your site. It reads what everyone else says about you. Quarterly review. Which clusters are pulling citations? Where are you still invisible? What do the conversion numbers say? Connecting visibility to pipelineThe objection we hear most: "How do I prove AI search is driving revenue?" Fair question. Attribution isn’t fully solved. There’s no Google Search Console equivalent for AI. ChatGPT’s free tier doesn’t even pass referrer data. GA4 doesn’t have a native “AI” channel, so you have to build one yourself. Here's how. In GA4, create a custom channel group using a regex filter on session source:
chatgpt.com|perplexity.ai|gemini.google.com|copilot.com|claude.ai
That gives you a floor. Visible AI referral traffic is estimated to represent only 30-40% of actual AI-driven visits, since many users see your brand in ChatGPT and then Google you directly. What we know about that traffic: it converts. A Seer Interactive case study of a B2B client found ChatGPT traffic converting at 15.9% and Perplexity at 10.5%, versus 1.76% for Google organic. Smaller sample, but the signal is consistent with what we see across our clients. Track AI-referred sessions, brand search volume trends, and third-party referral traffic from sources AI cites. If citations are increasing monthly and brand search volume is trending up, the pipeline impact follows, usually with a 4-8 week lag. This gets bigger than searchIn September 2025, OpenAI launched Instant Checkout. Users could complete purchases without leaving the chat. Walmart partnered with Google Gemini for the same thing: describe what you need, AI finds it, checks availability, handles checkout. Last month, OpenAI pulled back on Instant Checkout. So far, the shift to agentic purchasing is off to a slow start. But the protocols are being built, and the general direction is clear. And the projections are… wild. To say the least. Gartner estimates that by 2028, 90% of B2B buying will be intermediated by AI agents, pushing $15 trillion of B2B spend through agent-to-agent exchanges. McKinsey projects AI agents could orchestrate $3-5 trillion of consumer commerce by 2030. Already, 45% of consumers use AI for some part of their buying journey. This changes what AI visibility means. It's not just "Does AI recommend us when someone asks a question?" It's "does the AI agent choose us when it's autonomously making a purchase decision on behalf of a buyer?" The companies building visibility now aren’t just optimizing for today’s AI search. They’re training the models that will be making purchasing decisions. And that is where the real flywheel kicks in. An AI agent that’s learned to trust your brand through hundreds of citations, positive third-party signals, and consistent presence doesn’t start from scratch each time. The feedback loop we described in AI search becomes even more powerful when AI is making purchases, not just recommendations. Wrapping upWe started this series with a test: ask AI about your industry and see if you show up. If you weren't in the answer three weeks ago, the question is whether you will be three months from now. Because by then, the companies already doing this work will have three cycles of compounding data, real visibility across AI platforms, and a head start on agentic commerce. We built Saturation to run this system for companies who don't have the bandwidth to do AI search in-house. The recon, the content architecture, the distribution, the monitoring. If you want to see where you stand, start with a free AI search audit → Thanks y'all! |
|
Here's how we can help 👇
|
|
|
We get startups into AI search results.
|
Saturation
Free AI audit → |
|
We help startups build and scale growth engines.
|
Demand Curve
Work with us → |
|
Learn to build your own growth engine.
|
Growth Program
Join 4,500+ startups → |
|
Access the top 1.5% of global growth talent.
|
GrowthPair
Get paired → |
|
📣 Want to reach 105,000+ growth leaders? Sponsor this newsletter →
|
|
|
How'd you like this week's newsletter?
|
|
Click to vote — we'll use your feedback to improve the next one.
|
| 💙 This was great. Loved it! |
| 😵💫 Meh.. it was ok. |
| 😭 Not good. No Bueno. |
|
© 2026 Demand Curve Inc. All rights reserved. |
Doxxing Case Conviction
