Wil Reynolds's Recent LinkedIn Posts

The typical female pet owner spends more time actively engaged with her pet than she spends in face-to-face contact with friends of her own species. Men who watch television now spend 7 hours in front of the TV for every hour they spend hanging out with somebody outside their home. Since the early 2000s, the amount of time that Americans say they spend helping or caring for people outside their nuclear family has declined by more than a third. These data points hit, make you question, reflect, etc. I'm not here to fight the logic but to reflect and say do these things fit me and my lifestyle and if so, am I ok with that? https://lnkd.in/e7PSE9K7

There's so many ways to show GEO case studies that most buyers would fall for, be careful y'all. Here's two examples: First, starting December 1 our research flagged that typical citations went from 5.7 to 10.4 more citations. Longer answers could equal more brands showing up in answers, more brands = more visibility so your visibility could have gone up looking great in the case study but having low material impact on your business. Imagine if you were getting an SEO report for page 1 am suddenly Google changed to include 30 results on page 1. Did you page 1 visibility go up, yes. Your lead flow probably not. Second quick wins like low quality listicles could easily have given people a "visibility bump" shown in a case study, keep an eye on the dates in those case studies - because a study we just dropped showed listicles are having much less impact in the last few weeks across several industries. If the case study isn't fresh, then what worked back then may not be working so well now.

Making a strong case for distribution of your company information / news / etc in a GEO world. This part makes me think a lot about the black box of training data and the long horizons of training data cut offs. You combine that with the obvious proof that repetition works in AI for now, and it's like yeah get your stuff out there more in more places sooner than later. As the models go to higher quality sources of training your relationships with industry publications will matter even more to corroborate what you are saying about your brand. "When a story lives only on a brand’s site, the model has just one place to encounter it. When that same story appears across multiple reputable publisher domains through earned distribution, the model encounters it in more contexts, which strengthens the overall signal around the information. This broad, multi-domain presence increases the likelihood that the story will surface in AI-generated answers."

We analyzed 2M+ citations across prompts over 3 months - Listicles are finally taking a hit! (30% to be exact) But the hit isn't universal, check your industry, here's our findings... 📉 ChatGPT listicle citations decreased 30% (but check your industry in the post) 📈 Wikipedia = winner when listicles got hit (Wiki has a robust human approval process, looks like dialing up trust could be on the horizon). 🔢 Agency listicles = seemed to take a hit, but instead of wikipedia it is clutch.com moving the needle. High Citations = Higher listicle reliance - When ChatGPT had its most citations per response (12), we also saw listicles being referenced the most (19.4%). Maybe there is a precipice at which "more citations" doesn't help make "better answers"? What about Google AI Overviews? - Well in some industries listicles are actually increasing in their efficacy. Google seems to combat low quality listicles with "disclaimers" warning users while not "penalizing" the listicle builders. We also break down some "good" listicle characteristics in this article with 4 examples of companies doing it better than most. https://lnkd.in/ecmeWZjM

Every CEO / CFO should have a stack of gift cards in 10, 20, and 50 dollar increments so you don't get stuck paying for subscriptions team members may have stopped using and forgot to cancel. It doesn't scale that well, but it is good for smaller businesses who want to empower teams to "keep pushing and trying new things" but don't enjoy looking back at a credit card statement and seeing 8 or 15 monthly recurring charges that stopped being used 3-4 months ago. In mid sized companies, solutions like DIVVY cards are helpful (that is what I know we used).

There's so many ways to show GEO case studies that most buyers would fall for, be careful y'all. Here's two examples: First, starting December 1 our research flagged that typical citations went from 5.7 to 10.4 more citations. Longer answers could equal more brands showing up in answers, more brands = more visibility so your visibility could have gone up looking great in the case study but having low material impact on your business. Imagine if you were getting an SEO report for page 1 am suddenly Google changed to include 30 results on page 1. Did you page 1 visibility go up, yes. Your lead flow probably not. Second quick wins like low quality listicles could easily have given people a "visibility bump" shown in a case study, keep an eye on the dates in those case studies - because a study we just dropped showed listicles are having much less impact in the last few weeks across several industries. If the case study isn't fresh, then what worked back then may not be working so well now.

Really good read that is basic enough to give all of us a good framework to think about information retrieval in an AI world: How LLMs and RAG Systems Retrieve, Rank, and Cite Content.

Branded search is SUPER important to track, we had a single review posted on several sites impacting our branded prompts today, so this is what we are doing about it. In the immortal words of "B-rabbit" it's like tell them something they don't know about me, the marketing version of that is posting more about your company, and not just once in a blog post which we found to lose effectiveness, but a consistent place. Shout outs to Amanda Natividad and Harpreet Singh in this one. https://lnkd.in/djKybvFx

This post is a great read for all marketers, have you codified what AI is allowed to help you do vs what it completely does and thought through why? When is it doing work for you vs when is it relegated to a collaborator at best? Kathleen Booth really gets you thinking on this one.

Hey friends in Germany and Europe broadly, I'll be there to share my most recent thoughts on where do we go from here? AI disruption is real, we don't even know what to call our industry yet! I really hope to give y'all something to level up your thinking and your actions to win over the people and not just the prompts. Prost!!

Is the real moat for tech companies in being a default infrastructure recommendation in Open AI's Codex or Claude Opus 4.6? Hear me out on why... I have built 5-6 vibe coded solutions, 2 niche AI trackers, a personal CRM, a slide deck expander tool, etc etc. These tools are built on whatever the AI recommended when I was vibe coding. I didn't know what "Supabase" was 4 months ago. While vibe coding in Opus 4.5, it was recommended to help me build what I wanted, so I signed up, got an account and now I pay them 40/month. Same with Vercel. Yet when most enterprise infrastructure clients come to us at Seer - they are asking about ChatGPT & Gemini, and now that I am vibe coding, I can't help but wonder - shouldn't we track your visibility in Codex? Also in all my coding in Claude Code (4 months now), I'm realizing OH SNAP, it never searches the web, meaning all these new coders like me are likely to use whatever the training data says. This means getting into the training data is a MOAT because they just don't update the models training cut off dates that often. Then I realized - my entire tech stack was recommended for me by Claude Code's training data, and I never questioned one of the recommendations, and everything I wanted got built well! The other thing I think is so important, is for infrastructure and SaaS companies to find "where does the AI constantly get questions about how to troubleshoot wrong"? I almost left supabase due to a port issue because every time I ask Claude Opus 4.5 how to troubleshoot it, it never considers my personal network could be blocking supabase's ports. Same way in 4.6. Needless to say I almost left supabase because I spent 2 hours trying to get something working that the training data didn't know, luckily I googled it, got the answer and learned, but SaaS companies could lose customers not because of anything they did, but because the models haven't been nudged to know how to troubleshoot in certain cases. All this to say ... GEO could be a lot more than where do I "rank" in ChatGPT and AI overviews, especially in SaaS and Infrastructure - but a GEO firm could help you: 1 - optimize in training data as a typical default answer (i.e. every time I wanna text Twilio is always mentioned) 2 - find bad vibe coding "loops" where the major AI coding models, don't give good recommendations - hoping to help with "retention" Thoughts?

Getting way too many "here's this post on our tool, if you post to LinkedIn tell me what we can pay you for it" outreach. These people are the new "paid links" folks, just in new clothes. Crazy part is one of them emailed me, I was like donate 5k to this charity assuming they would leave, they came back with 1k. Then I said, if you actually look at my feed, I already shared one of your posts for FREE 3 weeks ago. Keep your eye out for 1 - "Influencers" sharing stuff from companies they never linked to before 2 - Other influencers sharing stuff at the same time to the same link 3 - Embargoes - after they post, these companies want the stuff to "stick" so the people who take the bribes, errr compensation, might dial down links to competitor info and studies for a week or two afterwards. There has to be a better way, right?

If you don't develop a KPI for trust, your content team will slip into AI slop mode, increasing production but decreasing brand trust. Of course there are things to takeaway that create value, but really dog? 8k for 30 bux? I keep seeing people talk about replacing full marketing teams with AI. My first question is always the same: what kinda low quality marketing were you doing? In the same way that "you have a pre-approved loan" spammers aren't marketers, they might be effective, it might work, but it isn't marketing. We're watching companies write 50 posts in a single day and we all know that doesn't build brand, but it keeps happening because output is easy to measure and trust just isn't. So if you're scaling with AI right now, ask yourself: do you have a real signal it's working? Not how much you produced. Are people trusting you more? Take a look at that "fast" content you built, what does social or direct traffic look like? The goal never should have been more. The goal is better. Here is the link to the dashboard you can copy to see if you are getting AI traffic and no social to your posts! https://lnkd.in/ebGbGHK4

A few days ago I shared the conundrum I was in of not wanting to post too much on LinkedIn, but I am at time where my brain is processing so much information that I am trying to work through - that I would like to have that outlet, this newsletter (first edition) is my attempt to drop more unsructured thoughts. Right now, this first post is forcing me to show it to al of you, I hope that in the future I can control that better. Cause sometimes I have 5 unstructured thoughts in a day and that would be spam for those if you who follow me but don't want the full firehose (and I don't blame you). I intend to start recording these too for my peeps who reocmmended podcast format. The goal is to NOT spend 10 minutes trying to make the right thumbnail, the right post, worrying about text length cutoffs, and preview windows, all the stuff that gets in the way. Sometimes I'll voice transcribe longer stuff, sometimes, it will be quick, like a link hat is making me think, or a song that I was loving that day. Welcome to the jungle.

If I posted today all the killer stuff I've curated for you guys, most of you will never see it. When Rand Fishkin posted about problems managing what to post given the proclivities of the linkedin algorithm, he shouted me out saying that largely Wil is a "IDGAF kinda guy", which I responded not so fast, I do play one part of the game, and it just happened. To maintain my focus I typically save things to read, and then I'll read 30-40 articles at a time. I just did that today. Now I have 10-15 ideas and I refuse to post them all as separate posts b/c of "the algorithm" and having to "space things out". This kinda sucks. Not for me, but for those of you who would rather get all the things, than some spread out over 3 - 4 weeks version. My goal after going through my reads is to share ONLY things I think that will help people think. So with that goal in mind, here are the options I am thinking of: 1 - I could tell everyone "hit the bell icon" (which you should) to make sure you don't miss ANYONEs ideas that you value on this platform - the amount to times I miss content from people I love to get some low quality update congratulating someone on a job whom I don't even know is insane. 2 - Post to the Seer website + newsletter, and spread out on here over time. Pro: existing infrastructure exists, no rented land. Con: the stuff here gets spread out, people who aren't subscribed miss out 3 - Make a linkedin newsletter (or article) that allows me to just DUMP and whoever sees it sees it. When I just wish it was easier for people to see what I am reading and my commentary on those reads. Pro: Gets (potentially) out to more people on this platform, meaning I should do the same on X too. Con: Managing another newsletter(s). 4 - I have a medium account, but I'm thinking about moving off the platform, every time I try to use readwise on a post it blocks it and what not, so I think they are a bit too much "thumb on the scale". Pro: Speed of posting / getting stuff out. Con: Having to do the work, creating broken links, duplicate content. Linkedin audience isn't there. So good content has to be pushed constantly here to get it to move over there. What do you think? Trying to figure out how I get more out when I have these bursts, where yall don't see most of it b/c I posted too much in too short of a timeframe. Love the network effects here, cause you can help more people think. Like Rand, I don't enjoy playing this part of the game at all. I don't care about link in comments and some of the other games, but the "if i post too much" game is getting old.

A talented photographer could take a better show with a 5 year old iPhone than I could with $20,000 worth of equipment. That one hit me... it's a theme you can expect me to bring into more presentations, why? Because we are obsessed with "tools" what tools are you using, what AI tools help you to do ABC task. Look around at the next conference you are at "10 tools to win AI" will be packed, people will take copious notes... we've become wired for "tools" give me something I can put my URL into so work can get done. This article from Ken Norton was a darn good reminder that yes tools have their place, but give me Serena William's tennis racket or Tiger Woods Clubs and shockingly you aren't becoming them. So that is a reminder, a tool in the hands of a professional is different than a tool in the hands of a student learning. https://lnkd.in/eqhKVPDp

Sunset in Munich. Beautiful. Recommended accurately by a single prompt in Gemini., so I walked over with Nora and took a great photo.

Today I learned that Barack Obama, during the would retreat alone to the Treaty Room to handwrite personal letters to ten ordinary American citizens every single night, a practice he maintained with almost monastic devotion across all eight years. If he can do almost every day while being president, you can probably do it once a month. Try tomorrow.

People are being asked "why aren't we showing up in AI" and already overworked marketing leaders need answers quick, unfortunately the sales people are like "this is new, I dunno" so they don't really know the detail of what they are selling and our industry will be worse for it, this post is my attempt to educate at scale, so please share it, question it, and aske me a TON of questions so everyone makes better buying decisions. Yes I am still learning, so let's figure this out together, if I get something wrong, I'll let you all know.

Concerned about AI indexability of your content? Hit F12 CTRL + Shift + P Type "Disable javascript" Reload your site I am pretty sure Pedro Dias is doing the lords work helping us non technical types feel empowered. I drop this in all my preso's to help marketers feel like they can run small indexability tests on their own. https://lnkd.in/dQRF4N_3

Stop letting the worst 1% of humanity fill 90% of your feeds. Re-training algorithms to show you the good in the world.

This is Joe, employee number one hired in 2005. He believes in me enough to take a job with no health insurance. I bumped into him today at Indy hall thx Alex Hillman . The gratitude you feel on employee number one is always a weird feeling, at least it is for me. He believed in me. No health insurance no recognizable client names, none of that. it's crazy to think that in this picture from 2007, you can see Joe there when we were 5 people, Crystal O'Neill is still with me, Patrick Tigue and I are tight AF and when I travel I go see him, when he lived in Costa rica I saw him, Portugal saw him. feeling blessed about the journey.

This is how you determine what % of your job could be replaced with AI in 2-3 years. I just got asked a very strategic SEO question by a client on domain consolidation. Step 1 - Recorded my answer based on my experience Step 2 - Pasted the client transcript into Claude Opus 4.6, Gemini 3 pro, & ChatGPT 5.2 Thinking, told each to give a recco Step 3 - Paste in my answer with the prompt "Give me a side by side comparison of what you came up with vs what I did, and who "wins" on recommendations, what did I miss, what did you get? This will be a fun challenge for us a collaborators to play to both of our strengths. Step 4 - Review the output, determine what % of my reccos were already "in the machine" which ones were not, by model. What % of my answers / approach was also surfaced "in the machine"? ChatGPT 5.2 = 35% Gemini 3 Pro = 15% Claude Opus = 20% At times one model would give me ways of thinking I saw deep value in. i.e. Claude was the only one that nailed an LLM angle I completely missed ChatGPT went all "consulty" on me going for exec preso vs building a testing methodology. It said: You optimized for truth via experiments. I optimized for avoiding the wrong decision via framing + brand-demand protection. Gemini defaulted to super tactical SEO, avoiding financial impact and how to test, it went to how to make an audit and how to execute - Gemini "flagged" that I missed overly tactical stuff, which I assumed were table stakes. Ultimately this is a reminder to test all models: Be honest with yourself about what % of your answers are in the machine Keep on testing as new models launch Whether I like it or not, 15-35 of my work was in the machine. That means I gotta focus on other areas, which one thing I would do they wouldn't have is internal data on prior consolidations, which we have available via our MCP, which can search slack and then go pull analytics data for any clients where we discussed domain migrations. Things you can do instead: Be cross divisional - none of these ever really mentioned paid search and the data of paid taking up real estate. Claude was the only one that said, "hey think about LLMs and brand defaults (which I, Gemini, and ChatGPT all missed). Stay up to date - Aleyda Solís just did a cool report on SERP real estate that I just read THIS weekend, I could look at changes in SERP real estate form her tests. Get your data normalized - The more you can combine data across clients the more you can use an LLM to pull that data and tell you what happened, find a way to do this in an LLM: search slack, find clients who migrated, then pull transcripts and reports to see impact. Ultimately, use all the models as a collaborator. I know we all hear that a LOT, but I wanted to show what that could look like. Note: This could be a good use case for Perplexity's new "dogpile" search.

If you know me, I downplay wealth a LOT, I'm uncomfortable talking about it, and I'd rather people not know anything about me other than I love my work and I get paid decently to do it. However when I went through the opportunity to sell, I realized no one was giving the rawest parts of what can go through someone's mind when "life changing money" is on the table. A recent post from someone made me reflect on this 10 year old post and update it.

Right now a spammer who has never spoken at a conference I've been at is CRUSHING us according to ChatGPT, Claude, Gemini, and Perplexity on GEO / AI thought leadership. I'm ok with that because that isn't the prompt to win on. Stop tracking: "Best GEO Agency for X" Start tracking: "I was recommended Seer Interactive and [Company B], I'm looking for these 3 things, give me a pros and cons, or comparison." Look at the differences, run it several times, what sticks, what doesn't. People ask people! 77% of B2B buyers ask their network first. The best buyers are educated and mature, they have networks. So those buyers might not just go to chatgpt being like give me a GEO agency, they are looking to take the brand names they got from trusted friends and compare. I built a little tool for fun to show you how you line up on a set of attributes, and opened it for everyone to copy on github (it is my first time, so take it easy on me), hoping to make it better for the community. https://lnkd.in/dQxupRKn

I got some great questions on this, like "Wil, what do you recommend a web analyst do if their team can self serve a part of the job they used to do?" Here is an answer: I am playing around with agentic browsing right now, and it hit me. 1 - Run agents against YOUR web site to start tasks using agentic browsers like chrome and comet. 2 - Watch the patterns, you can see the task summary text - re run the tasks multiple times, do agentic browsers always scroll all the way, do they always search first? 3 - Go run that task against real users - do real users have different patterns? If they do, now you might be able to identify patterns that are likely to be agentic browsers, which is a thing on a LOT of C-levels minds - will I even need a website with images / upsells / etc in the future? 4 - Now you can show (don't send an article, a blog post or an opinion), show them that you are testing this with 5-10% of your time and take those hypotheses and track them monthly with some kind of a "when to care" threshold. This is the kind of work you can do when parts of your jobs get automated. How do I think about things like this? I read stuff, a LOT of stuff, a LOT of broad topics, and it gets me thinking in new ways.