Lennart Vallo

If you're a solo grinding SaaS vibe coder, they are lying to you You wont EVER get traction to your product without some paid marketing This comes from someone who makes his living off getting 'organic traffic' to established businesses

@GrindstoneSEO ohhhh sir i see i haven't told you about dispatch yet 😳 sorry to ruin your retreat

This plus a timer on my phone to come check 4 Claude instances every 15-20 minutes.

@foley_seo SEO is the bicep, GEO is the tricep. They're the same arm and more gains is ALWAYS better on both 😅

SEO is DEAD. Dead fun. There's still a HUGE opportunity for SEO in 2026. Create amazing content without being lazy. Put more effort into creating media to go with text. Make sure you've got a solid website architecture. Get and earn amazing links. Distribute content via social. Get better engagement. There is SO much opportunity for more clients, customers, sales via traditional search. Whilst the GEO community is scrapping over sub 5% click-share, I'm here enjoying those organic gains. #seo

@lilyraynyc @ahrefs Hot take: Reddit is overrated for AI visibility 😳

Thought I was seeing a lot more Reddit in AI Overviews and yup. They’ve been showing up more in there for a couple months now: (@ahrefs brand radar data)

Imo a good SEO in the age of AI not only optimize for getting high quality WARM leads through dominating the narrative He also considers on site engagement. You can generate 10 lead magnet ideas in 3 minutes, run eval against them and a/b test. (tools can get cited by AI too)

Yes it's a good point i love how intellectually stimulating it is to work with ai visibility vs trad seo. Each case is truly different and there is no universal playbook. In terms of your example here i would argue that multi location business in general would make up for less hyper specific regional juice with better and broader authority signals. Obv not always the case but in general. They would also cover more surface for higher training cycle inclusion probability.

with ai now though, the fix isn't free like that. updating the about page, linkedin, press kit, schema, and bios to reflect a distributed footprint will improve AI surfacing — but it also dilutes the primary-city signal that's currently doing massive work in your local pack rankings and organic primary city targeting. the brands taking the worst hit are the ones with strong local SEO compounding in one metro, because the same framing that made them local SEO winners is now making them invisible to AI in the cities they're trying to expand into. there's no clean fix. for now you're choosing which lens to optimize for, and the in-between is the squeeze

reading my own breakdown of a multi-location agency client and it dawned on me: multi-location businesses might have a harder time with AI visibility than they did with SEO Google/SEO can follow the bouncing ball. AI's prefer neat packaged info...

Yea it's both brands / businesses and SEOs that need to adapt to this new reality though. A lot of prospects i talk to on the daily feel uncomfortable with the unquantifiable nature of this work. Most realize 'AI is important' at this point but there is still some gaps that needs closing when they're used to their marketing spent, not only SEO but paid ads as well have instant feedback and dashboards. There really is no ROAS for this game yet

yes this makes sense // seems accurate. you could say what SEOs built their careers on wasn’t really “do the work” - it was “prove the work” via a dashboard tight enough to defend budget every quarter. that loop is breaking / has broken and the anxiety is downstream of the measurement collapse, but the work itself is still legit. this is actually a return to how brand-building, PR, and content marketing have always worked outside of digital. you spend, you wait, you measure aggregate brand health quarterly, you trust the cumulative effect produces commercial outcomes you can sometimes attribute and often can’t. SEO had a ~20-year vacation from that reality because the measurement chain was so clean but I think the vacation is ending your “ton of effort → ??? → compounded performance over time” is exactly the new operating model. and i think the discipline is going to stratify around it. basically has to. the practitioners who can sell the compounding curve to their stakeholders will pull away. the ones who built their identity on the dashboard will struggle until they make peace with the slower clock or move to a different kind of work SEO has died many times. This is just a reset

love me some of them ahrefs data shares. agree the data shows what it shows bc paid IS rising and that's a real squeeze. one piece worth adding/some missing context though: panel data only sees web referrals. per @ethan_graphite's team + @Similarweb (Mar 2026): 83% of AI usage is mobile apps. 4-5x of actual AI usage doesn't show up in any web panel because it never produces a referral. the user opens the app, gets the answer, closes the app. zero web traffic generated, lots of usage happening, but not a part of this data. so the 0.26% is accurate for referral share, but AI is actually 56% the size of search worldwide when you count what's happening inside the mobile apps. the squeeze here has three sources imo, not one: paid, zero-click web AI, and mobile app dark funnel. x.com/timsoulo/statu…

Optimizing for AI authority around a cluster of topics is a very different game than classic SEO. The cozy and quantifiable feedback loop of the past with Impressions -> CTR -> Engagement -> Conversion is dead. Be comfortable being uncomfortable

@GrindstoneSEO @brucebiz2 @myeyesshine_ Ya ikr i would save so much money if it wasn't for the 2 mac mini's, the firecrawl, the 2 claude max subs, the claude api, the codex sub, the perplex api, the xapi, the openai api and lord knows what else that bills me every couple hours

@wearecited @brucebiz2 @myeyesshine_ Yeah, I basically on have data sources and APIs now. Everything else hit the cutting room floor.

I have sites Google refuses to index beyond the home page yet they're getting more than 100% coverage rate in GPT across a huge query fan. Careful who you read, the AIEO landscape is worse than the blackhat SEO landscape of ~20 years ago.

@ariozick @GrindstoneSEO @myeyesshine_ @brucebiz2 Ya training layer inclusion is obv the holy grail for now

@wearecited @GrindstoneSEO @myeyesshine_ @brucebiz2 I imagine it's useful for l2/l3, seeing who is used as sources etc and then getting placements there.. last week I finally understood why people have been emailing to be on best of pages that don't rank for much

add moar promptz to the prompt set, cross test citation share / tracking against g4a refs and dark traffic estimates is the best we can do here i believe .. Unfortunately chat terminal conversation history is not open source 😭 These damn tech giants and their respect of privacy amirite

@GrindstoneSEO @myeyesshine_ @wearecited @brucebiz2 The big question is how much does this help once the personalization on the front end? Meaning your dashboard makes the assumption that api output = client output and I don't know if that's the same thing

Perplex doesn't really have 'training layer' the same way GPT has, its whole system is build around live search, ofc they feed their models with base knowledge but it can't really be measured i dont think.. I think its base layer is pretty heavily influenced by GPT anyway so if you get in that training layer you should be pretty well covered

@wearecited @myeyesshine_ @brucebiz2 This. Haven't figured out how to do it for Perplexity yet though so it's reporting that score on 3/4 providers.

@myeyesshine_ @GrindstoneSEO @brucebiz2 mentions where a web search was not triggered i assume?

@GrindstoneSEO @brucebiz2 @wearecited 👀 cool to see L2/L3/L4 in actual product UI. curious what you’re using as the input signal for the L2 training score specifically. that’s the hardest one to measure because it’s inferred from outputs rather than direct.

@GrindstoneSEO @brucebiz2 @myeyesshine_ looks solid, if only all the pre 2023 SEO subscription dashboard services were a stock they would be the easiest short of a lifetime !

@brucebiz2 @myeyesshine_ @wearecited This is what mine looks like currently but I'm about to put in a good session with some logic upgrades/enhancements right now

How to actually start using AI for SEO/AEO superpowers: 1) DON'T USE THE APPS, not the chat web app, not the code web app and not the desktop app. Use your subscription to auth and install claude, codex or whatever model in your terminal. I know it looks scary but its plain English, get used to it. If you are scared i will allow you to have the chat web app guide you step by step on how to install in your terminal, simply copy paste. Another decent transition tool for complete beginners is VS code with the claude extension, there you can have the chat interface that you feel comfy with and a terminal instance running underneath. Have the chat instance guide you through the terminal and slowly transition into it overtime. 2) Don't copy paste skills blindly from GH repos or marketplaces, they're all generic crap and harms more than helps. They come with a bunch of fulff and constraints. Don't copy paste generic guru prompts you read off X, they offer you no value and just produced probabilistic outcomes of letters and words the engine think you would want to read. Claude, opus 4.6 is AMAZING out of the box, let it rip and collaborate with it to build your own skills and knowledge files over time. 3) When you finally get comfy with the terminal start working with claude to build your claude.md (you chat in plain english and have claude actually write it, you read and approve), this the bible of your project. Keep it lean 100-300 lines, couple lines with broad context of your project then rest points to your skills and knowledge files. Every time you open claude code the instance you chat with gets all its context from here. Trust me, having a lean and clean claude md from the get go will save you a ton of headache and keep the agent in line. 4) By now you have a somewhat decent foundation to work with and its time to experiment. With a 200$ claude max subscription you literally don't need much more, cancel all your stupid pre 2023 SEO dashboard subscriptions. Start with web fetch which is a free inbuild plugin for claude code, does the job fine for now. Now you can let your imagination run wild, depending on your experience level in SEO you can start off a couple different paths. IMO ironically enough the LAST part you should experiment with is actual content production, this is the most difficult to get right. I have done nothing else but building with claude and codex for aeo and seo purposes for 8 months straight 12 hours a day and i have still not solved it to a place where i feel like its good enough to just let rip and walk away. My system has a 74 point review checklist, excluding cross model council reviews, adversary reviews, 100 point scoring system humanizer review and a final pangram api check. STILL i iterate on content and proof read. 5) Research, this is where AI shines, with free tools like GSC and G4A hooked up (have claude guide you on the setup), claude in chrome free plugin you can go to work, if you're a beginner have claude implement proper tracking on your site, let it identify gaps in your metas, let it look for optimization opportunities across your site etc etc etc, with a little guidance its amazing at this. If you're an experienced SEO correct and steer it a bit as needed and have it write it down into skill files over time so that you can template your workflow. This is essentially the best use of these tools when you have domain knowledge, slowly describe your workflows in A -> B -> C format and take your hands more and more away from the keyboard as you begin to trust the processes. If you want a tool in your kit i think might be worth it right away for SEO purposes i think KE is ok, its cheap and provides value, talk to your claude code about it and have it guide you - (for best practice save a txt file .env and copy paste the api key in there and have claude pull it, dont leave api keys in chat's stored on anthropic cloud jsut for good opsec.) 6) For querying AI and optimizing for AI visibility: The end game is api keys to all the models and sorted data in json files with a set monitoring cadence and an ever evolving prompt set that grows with your onsite and offsite efforts but in the early days just start with claude in chrome in an incognito browser and have it query gpt free, gemini, perplexity and AIO - have it log the recommendations and citations so that you can track it over time. This is a bit slow and messy plus browser control will cost you a bunch of tokens but its free (ish) and ok when you first start out. 7) By now you have build yourself an evolving system that knows your workflows, gets smarter and fine tuned over time, builds proprietary data sets via YOUR AI queries and YOUR gsc/g4a data. This is your moat and your path to actually win and evolve over time. 8) If you get lured in by new exciting stuff you read on X, copy paste the article link ask claude how it suits into YOUR system, if you find claude agreeing right away ask it to run an adversary agent (first mention of a sub agent) against implementing and weigh that report. I advice you to always be really cautious polluting your workflows with new shiny tech. Knowledge and context blows up fast and can mess up beyond fixing if taken to the extreme. 9) Final thoughts: Understand that you can now control and direct intelligence with plain english. Dont be scared go WILD on the experiments but at the same time be very cautious of what you actually save into skills and knowledge files. Slowly add workflows and be rigorous with the testing. I have a ton of experimental sessions where i try out a bunch of wild ideas but i always make clear that nothing gets implemented into the system without my approval after proper testing. The most important thing is that you have full overview and control of your process, system, structure and workflow guard what goes in here heavily. AI will tell you how brilliant you are do stuff in minutes that took you days and its easy to get lost in the sauce, second guess it, you are not jesus and the AI is only as smart as its system and its CEO (you). Peace

@GrindstoneSEO @brucebiz2 @myeyesshine_ If i write a comprehensive post with tips, structures and workflows i actually use do i get a repost :)? I promised myself i wouldn't engagement bait though, my eyes cant stand another COPY THESE 3 PROMPTS AND TURN BILLIONAIRE IN 3 WEEKS 😅

@brucebiz2 Just tell Claude what you want to build and it'll do it. If you want to get up to speed with the 'how this all works', there's way smarter people on here than me worth reading. @myeyesshine_ & @wearecited have both posted a lot of stuff that I ingested.

SEO impact on visibility is soooo query type, market, industry and GPT model dependent. The free model which majority of people use is so cheap with websearch and relies heavily on training data fx there is huge variance here but i will push back on the OP saying its all SEO, that is just plain false

@KaiCromwell scarcity mindset alert

98% of the revenue you’ll make from SEO in the next 12 months will come from Google. The remaining 2% will come from LLMs. Brands that are scrapping traditional SEO for AI SEO are conservatively throwing away 12 months of organic revenue on a channel that’s not even fully established yet.

@TTrimoreau Google doing A LOT to desperately hold tight on their monopoly i think we'll see divergence, different engines for different preferences but for the time being i think Google AIO will remain king for 2026 in terms of broader discovery

Which AI tool is replacing Google in 2026? 1.Perplexity 2.ChatGPT 3.Claude 4.Gemini 5.Mistral AI Drop your pick