Market Brew (@mktbrew) header
Market Brew (@mktbrew) profile picture

AI Visibility Isn’t Luck. It’s Signal Alignment. Patented search models, PSO and template models. Enterprise software with white-glove guidance.

Joined September 2014
Photo by Market Brew (@mktbrew)Photo by Market Brew (@mktbrew)Photo by Market Brew (@mktbrew)Photo by Market Brew (@mktbrew)Photo by Market Brew (@mktbrew)Photo by Market Brew (@mktbrew)Photo by Market Brew (@mktbrew)Photo by Market Brew (@mktbrew)Photo by Market Brew (@mktbrew)Photo by Market Brew (@mktbrew)
Market Brew has achieved RSI: Recursive Self-Improvement for AI visibility. As Co-Founder and CTO of Market Brew, I have spent years working toward this moment. It can now analyze how AI systems understand a brand, identify why competitors are being surfaced instead, prescribe the highest-impact changes, measure the outcome, and use those results to improve the next cycle. Every crawl makes the model smarter. Every test sharpens the blueprint. Every result improves the next recommendation. This feels like an arrival. AI visibility is entering a compounding phase, and only a small group of companies will begin building that advantage early enough to shape their category. We are quietly opening Market Brew to those companies now. If you want to be a part of this rollout, our demo form is on the marketbrew [dot] ai site. See you inside.
1
1
1
2
235
Market Brew retweeted
Your customers might not see your cookie preference content. That doesn’t mean machines don’t. We recently analyzed GEICO's website in Market Brew and noticed something strange: cookie preferences were its third-most-important website cluster. Not exactly what you'd expect from an insurance company.
1
4
73
Market Brew retweeted
Replying to @mktbrew
The tool is especially useful for recommendations around improving link flow distribution. I had ChatGPT compare what it could do with a general OKF and Market Brew's OKF, and there's such a higher level of analysis on the MB one.
2
3
112
Replying to @crawler_gill
Well, basically we have millions of pieces of content. We only want to surface unique content to our users. So how do we do that? In the past, we would do things like: 1. compare keyword overlap (2003-2010) 2. compare topic clusters (2010-2017) Most recently, technology has allowed us to convert content to geometry (embeddings) which places each chunk of content into a coordinate system. Which means that we can now pick 5-10 centroid "points" that we want to surface to the user. Those millions of pieces of content all look the same geometrically. Unless you have, in your words, non-commodity content (which is a easy way to explain a page having a "unique centroid"). Essentially we are filtering out millions of pages to uncover 5-10 unique coordinates and then surfacing those pages. So the game is to make your content geometrically unique. Its a mathematical problem. There are tons of ways to do this, starting with creating a unique digital footprint for your brand, and anchoring your content around that.
1
1
8
185
Standing out in AI search ≠ rankings @scott_stouffer breaks it down: → Meaning > pages → “Different” ≠ different to AI → SEO = full-system alignment Start thinking in vectors 🎧 seoweek.org/scott-stouffer-2… 🎟️ seoweek.org/#pricing-1
2
2
2
5
355
🎤 Speaker announcement @scott_stouffer is joining SEO Week 2026 Founder of @mktbrew, he builds AI that models search engines and predicts rankings before you publish Excited to hear his take on predictive SEO 📍 NYC | 📅 Apr 27–30 🎟️ seoweek.org/#pricing-1
3
4
1
5
550
People say SEO is unpredictable. From the outside, I understand why it feels that way. Rankings move. Updates roll out. AI reshapes the interface. It can look like chaos. But as someone who has spent years building search engine models, I can tell you this: What feels unpredictable is usually just unmodeled. If you don’t understand: – where authority is bleeding out of your graph – how your pages cluster semantically at the embedding level – whether your internal link structure reinforces or fragments topical focus – which intent classes your competitors already dominate then yes. performance will feel random. Because you’re making changes without a mental model of the system evaluating you. Search engines are not mystical. They are layered systems: • link-based authority propagation • document classification and clustering • passage-level embedding retrieval • query fan-out and intent filtering • behavioral reinforcement signals If you optimize without understanding how those layers interact, you’re guessing. When you start modeling how a search engine is likely interpreting your site, how your centroid aligns (or doesn’t) with query space, how authority flows between clusters, how competitors occupy adjacent embedding neighborhoods... things stop feeling chaotic. Not easy. But measurable. Testable. Explainable. The difference between “SEO is unpredictable” and “SEO is complex” is whether you’re willing to model the system underneath it. Search isn’t random. It’s just mathematics most people never see.
People say SEO is unpredictable. But that's not necessarily true true. What’s unpredictable is trying to optimize without understanding the system underneath it. If you don’t know: – where your authority is leaking – how your content semantically overlaps – what gaps competitors have already filled Then yeah, it’s going to feel random. But when you start modeling how search engines are likely interpreting your site, things stop feeling chaotic. Not easy. Just less mysterious.
2
3
12
474
Different ranks. 🔎📈
1
3
140
.@sengineland just announced that I won Search Marketer of the Year for a second time. We don't get the opportunity to give speeches for things like this anymore, but this is what I would say if we did: First off: thank you. To the judges, to the community, to my team at @iPullRankAgency ...the most brilliant, relentless crew in the game. Winning this once was a statement. Winning it twice? That feels like legacy in motion. I didn’t get here by playing safe. I got here by breaking things open. By questioning every assumption this industry ever handed me. SEO was never meant to be static. It was never about chasing 10 blue links. It was always about decoding how meaning moves through machines. While everyone else was arguing over subdomains versus subdirectories and and what to call AI Search, we were out here defining Relevance Engineering and building the bridge between AI and human intention. We weren't getting lucky, we were keeping our heads down and getting to work. This moment isn’t just about me, it’s about what comes next. Because the future doesn’t wait. It doesn’t care who’s comfortable and cautious. It rewards the curious, the creative, the ones crazy enough to believe that search can be rewritten like code and poetry at the same time. So to everyone watching: stop playing defense. Don’t wait for the next AI innovation or algorithm update to tell you who you are. Be the signal, not the noise. The machines are learning from us. Let’s give them something worth learning. Thank you and congratulations to all my fellow nominees and everyone in this space pushing the boundaries. The future’s already watching. searchengineland.com/search-…
22
6
2
85
5,699
Google’s embedding systems already treat “easy-to-replicate” content as duplicates of meaning. The only way to stand out is by making your AI output brand-unique and search-modeled: 🧠 3 keys to break the AI-sameness trap: 1️⃣ Brand Bible: teach the model your voice, values, and banned phrases. 2️⃣ Template URL: give AI the structure that already wins in your niche. 3️⃣ Reverse-engineered Fan-Out Prompts: expand semantic coverage the way your competitors rank. The future isn’t “more AI content.” It’s smarter context engineering. 👉 brew.marketbrew.ai/login.htm…
Think you can win with generic AI content? Think again. The Google leak reveals OriginalContentScore & contentEffort signals. Google is literally measuring how hard your content is to replicate. If a machine can easily reproduce it, it won't rank.
1
1
3
203
🧵 SEMrush’s Fall & The End of Generic SEO /1 SEMrush’s stock is down over 50% in six months. Guidance slashed. EPS flipped negative. And this isn’t just a market correction... it’s a signal. The SEO tool era that powered the last decade… is dying. Here’s why 🧵
SEMRush stock has been on a consistent decline all year, even with it's pivot to AI and enterprise, and it's a damning tale... One that I am considering shorting 😅 Wall Street once loved SEMRush! It was a nimble SaaS riding the SEO boom while big tech focused on their ad monopolies... Fast forward to 2025 and the stock is down over 50% in six months, guidance was slashed, and is now an injured unicorn simmering on the edge of a sub-billion dollar valuation. This is all because of a handful of factors: - There core SMB and freelancer base, the lifeblood of SEMRush’s early growth, is bleeding out, and that's an industry wide problem. - The enterprise pivot? Still too early, too expensive, and not compelling enough. Large marketing teams already sit inside Adobe, Salesforce, or HubSpot ecosystems... - They bloated out the platform with too many mediocre tools, becoming a jack of all trades, master of none. Meanwhile, the macro reality for search is brutal: CTR from organic is collapsing as Google continues rolling out AI Overviews, AI Mode and pushing zero click experiences whilst publicly saying they will actually increase clicks somehow. Even if SEMRush builds the best tools in the world, they’re chasing a shrinking TAM. And the market’s noticed. CAC is rising, LTV is dropping, and EPS just flipped negative (-$0.04 when investors expected positive). Unless SEMRush can reinvent itself as a data infrastructure layer powering next gen marketing agents, not just another SEO SaaS, this decline isn’t cyclical, it’s existential. AI hasn’t just disrupted search. It’s rewritten the entire value chain of how marketing intelligence is created, analyzed, and acted on. And if SEMRush can’t evolve faster than the algorithms it once measured… Then this chart might not be a dip, it might be a preview. 📉
1
1
4
336
What if you didn’t need to pay for every click? 👀 One agency replaced PPC with Predictable SEO. And saw 5x more traffic and 3x less budget spend. No ad spend. Just precision. Curious how they pulled it off? 👉hubs.li/Q03BQn100
4
4
3
9
1,408
Replying to @GaelBreton
Absolutely appreciate the honesty and experience in your post. But here’s a hard counter from someone who’s not just speculating, but simulating search behavior at scale. LLMs are not the end of SEO. They’re the beginning of a new level of precision marketing. False Premise: “AIO/GEO is not intricate, it’s just based on pages LLMs scrape.” Not true. If you’ve actually analyzed LLM behavior in the wild (not just observed results), you'd see: It’s not just top-ranking pages that get scraped. LLM responses are heavily driven by semantic clusters, embedding overlap, and passage-based matching, not just URL prominence. Tools like @mktbrew have shown that tuning a site's content at the embedding level directly impacts AI Overview visibility. The "No ROI" Argument Misses the Bigger Picture Sure, basic info queries are down in CTR. That’s been true since featured snippets and knowledge panels. But AI Overviews bring something new: You can influence multi-passage inclusion. You can test, simulate, and win with precision, not guesswork. SEO isn't dead. Lazy SEO is. The era of “just publish content and pray” is over. The new game is technical, and yes, some of us are thriving. “LLMs are a straight nerf” ... only if you don’t evolve Every algorithm update in Google history took things away and gave others a chance to rise. Panda, Penguin, Hummingbird...same cycle. Winners adapted. Now it’s LLMs. Same rules: Test Learn Adapt The only real "shadow ban" is staying stuck in 2021 tactics. Social Media as the Alternative? Good luck building consistent business intent traffic there. Social is rented space. You don’t control discoverability... platforms do. Organic SEO, even in AIO, still allows earned, persistent visibility. Social burns fast. The Playbook Now: 1. Model your site like Google does, not guess. 2. Align your embeddings to the intent clusters of your niche. 3. Use tools that reverse-engineer what LLMs are actually pulling from, not just what ranks in traditional SERPs. 4. Rinse and repeat...weekly. Not yearly. Summary: SEO is harder, not dead. AI Overviews are gameable...but not in the cheap way. LLMs opened a new door for those who understand how they work. And that is the opportunity everyone sleeping on this will miss. Welcome to SEO 2.0.
1
1
5
611
See What AI Sees: AI Mode Killed the Old SEO Playbook — Here’s the New One via @sejournal, @mktbrew Google’s AI Mode is reshaping SERP rankings. Discover how embedding vectors, not keywords, now determine visibility. searchenginejournal.com/ai-m…
1
1
2
78
Study: Google AI Mode Shows 91% URL Change Across Repeat Searches. via @MattGSouthern: hubs.li/Q03vyrVl0 #Google #SEO
3
22
3
45
6,304
Market Brew retweeted
🧵 Google’s AI Mode is reshaping SERP rankings Discover how embedding vectors, not keywords, now determine visibility. Dive in!👇
2
4
9
1,152
Replying to @BrandLongo
Great thread — this is exactly why our reasonable surfer model @mktbrew adjusts link weights based on context, position, reciprocal, site-wide reciprocal behavior. We’ve even extended it with a PageRank sensitivity test (like US8250114B2) to flag inflated authority. If your team’s still relying on raw link counts or legacy authority scores, it might be time for an upgrade. teapawt.org/scott_stouffer/status/…
🔥 This upgrade has been a long time coming. For years, we’ve known the future of link scoring wasn’t just about position or prominence...it was about meaning. This new semantic link scoring system closes the loop: we’re not just simulating the Reasonable Surfer...we’re now embedding why the surfer chooses certain links. 💡 The big shift? We’re comparing anchor text vectors directly against the destination’s top thematic cluster, not the whole page. That’s a game-changer for precision. This is how you bridge link graphs with AI-driven understanding. Proud of the team for pushing this frontier forward. Onward 🚀
1
1
198
🔥 This upgrade has been a long time coming. For years, we’ve known the future of link scoring wasn’t just about position or prominence...it was about meaning. This new semantic link scoring system closes the loop: we’re not just simulating the Reasonable Surfer...we’re now embedding why the surfer chooses certain links. 💡 The big shift? We’re comparing anchor text vectors directly against the destination’s top thematic cluster, not the whole page. That’s a game-changer for precision. This is how you bridge link graphs with AI-driven understanding. Proud of the team for pushing this frontier forward. Onward 🚀
🧵 THREAD: MAJOR LINK SCORING UPDATE 🚨🔗 We’ve just rolled out a big upgrade to how links are scored—moving deeper into the Reasonable Surfer model with semantic link analysis. Here's what's new and why it matters 👇
1
2
136
More examples of how when AI Mode cites your business, it usually uses your GBP page; not a link to your corporate website. Worth checking in on that default GBP image & what you can do to influence it for optimal UX.
3
7
1
47
3,572