About
Core focus: demand gen, marketing automation, conversion optimization, paid acquisition…
Activity
14K followers
-
Kevin White posted thisI’m an agent-pilled marketing lead at an AI company. Here’s everything I built with AI in the last 30 days: 🥞 These are stack ranked by my own personal joy level: 1) AI search brand audit for prospecting Generates an AI search benchmark report using the Scrunch API. We drop in a domain, let it cook for ~a week, and it spits out a report that shows competitor stack rank, share-of-voice, and citations trends. (Sidebar: this one’s a personal favorite because it was my first foray into using APIs to build a tool) ––––––– 2) Weekly 15:5 update automation I h8 writing weekly updates. For context, a “15:5” is supposed to take 15 mins to write, 5 to read (but never does). After writing two updates by hand, I gave the task to Cowork: examples + instructions + data access. Now it writes 90% of it and takes about 5 mins to edit. (maybe now I should call it the 5:5 ehh?) ––––––– 3) Data analysis from Wikipedia citations I can’t disclose full details on this one yet 🤐 But we took a dataset of AI responses citing Wikipedia and ran analysis using the Wikimedia API—no data scientist required. ––––––– 4) Microsite for AI search products Vibe-coded a search-optimized microsite with Claude Code that compares products in the AI search (aka AEO/GEO) space. ––––––– 5) Cron job for attribution automation Setup a cron job that pulls from: self-reported attribution, call transcripts, and email exchanges. It then classifies + updates opportunity source in our CRM automatically. ––––––– 6) Lead enrichment (from merely title + company) We got a list of conf attendees with no PII data. Was able to take that list and have Claude Code give its best guess at filling in person + LinkedIn URL in a spreadsheet. It was ~50% accurate. Better than nothing. ––––––– 7) Guide for LinkedIn document ad I’m a little ashamed of this one because I’ve talked a lot of trash on white papers/ guides. But I got convinced to give it a try as an experiment. So I generated a PDF using a handful of blog posts and LI posts as the input. It’s also designed to spec using AI + brand guidelines. (Note: quality input is key here to not get AI slop out the other end.) TBD on the efficacy of this asset format. ======== That’s mostly it. I also chat with LLMs quite a bit as a sounding board, but I didn’t include that since it’s not exactly agentic work. What’s pretty wild is that, in the past, it would have taken weeks and cross-functional support. And, because of that, they wouldn’t make it past the idea phase. Now they can all be done with a one person with a hair-brained idea + an agent. And even when a task takes hours/days, most of the work is happening on background. P.S. Happy to share more details if anyone is interested. DM me and I’ll share the receipts on any of the above.
-
Kevin White shared thisWe just shipped a feature that helps customers pay us less. AI search vendors (Scrunch included) collect revenue based on a consumption model. The unit of consumption: prompts tracked. The logic goes something like this: the more prompts you track, the more value you theoretically get, and therefore, the more you pay. Great in theory. But not always in practice. At a certain point: -- New prompts stop adding new information -- New prompts return the same citations -- New prompts add noise, not signal And so we built a feature to show you where that point is. (I’ve been referring to it as the Goldilocks zone from prompts) It’s called: Topic Prompt Optimizer. It looks at all the prompts & citations within a topic and shows: ✦ Which prompts add differentiated value ✦ Which prompts are redundant ✦ Which prompts you can cut There’s a lot of data science work behind the scenes. But under the hood, it’s doing something simple: It measures how much new coverage (i.e., unique citations) each prompt adds… and how often the same citations show up across 2+ prompts. We take your full prompt set, run the analysis, and map the tradeoff. You see exactly where adding prompts stops adding new coverage and starts creating overlap. If you’ve ever wondered, “what’s the optimal set of prompts for this topic?” this gives you a clear, scientific answer. Demo below 👇 Or go deeper on our blog (it’s the featured post today)
-
Kevin White shared thisWhat’s the half-life of an AI citation? Answer: shorter than you’d think. After analyzing 3.5M AI citation events, here’s what we found 👇 ✦ Your avg AI citation loses ~50% of its presence in ~4–5 weeks ✦ Perplexity citation half-life lasts the longest (5.8 weeks) while ChatGPT turns over citations the fastest (3.4 weeks) ✦ Citation duration by industry doesn’t fluctuate much, all within a 4-5 week range ✦ Citations from established news outlets last ~2x longer (shoutout to Stacker for the analysis assist) Full data study is now up on the Scrunch blog (you know the drill). ==== Now, for some fun tinfoil takes by Kevin: AI platforms reward fresh content. ⤷ Not such a controversial take, but the short citation lifespans suggest that LLMs are constantly rotating in newer sources and phasing out older ones. Citation half-life may be a proxy for compute strategy. ⤷ And this gets into tinfoil hat territory: The reason why ChatGPT citations turn over in ~3.4 weeks vs. ~5.8 weeks on Perplexity is that OpenAI can afford to refresh aggressively, while Perplexity may hold citations longer to conserve compute. AI citations are rented, not owned. ⤷ Staying cited isn’t just publishing more and pushing out content slop. Unlike the SEO days of using the skyscraper technique, AI rewards reinforcing your presence across your site and the trusted third-party sources it pulls from.
-
Kevin White shared thisFun fact: I started my marketing career in ecommerce (shoutout Huntington Surf & Sport 🤙) Back then (circa 2000), the art and science of SEO was still in its early stages. Connecting the dots of a product catalog to SERP results was quite tedious. For example, I vividly remember the pains of: -- Generating XML product feeds -- Debugging those product feeds -- Tracking rankings via spreadsheet -- Tinkering with match types in AdWords -- Checking on competitor rankings w/ SpyFu -- Hand-optimizing pages for specific products Sidebar: It’s bringing back memories of the 🌅 AI Rising Sun boardshort craze which was a cash cow for us at the surf shop. It was messy, fragile, manual work that left way too much room for human error. It feels analogous to where product-level visibility in AI search is today. Tracking which products show up, where they’re positioned, and who’s winning takes a similar kind of manual effort. Not only that, it’s happening across many AI models (not just Google search like the past). Which is why we’ve made major upgrades to Shopping coverage in Scrunch. Instead of tracking products one by one, Scrunch automatically captures and analyzes product visibility, market share, and competitors across AI platforms for you. Here’s what it does out of the box: ✦ Detects which products make AI’s shortlist and which don’t ✦ Maps which products win proverbial shelf space in AI answers ✦ Tracks where AI sends buyers (your site or 3rd-party retailer) ✦ Surfaces the prompts you win and the ones competitors own It’s not enough to track AI visibility at the brand level when LLM recommendations happen at the product level. Scrunch lets you do both. Automatically. Shopping in Scrunch is now available to all brands with many products. Go deeper on our blog and/or watch my walkthrough 👇
-
Kevin White posted thisPro tip for anyone looking to land a new job: put AI somewhere on your portfolio. Literally anywhere: -- As a skill -- In the headline -- Update your job title -- Ask for an endorsement -- Bullets in your role description -- Post about how you’re using AI … You get the picture. AI is absent from 99% of profiles I review. But it’s the #1 skill I’m looking for in EVERY candidate. If you show how you use AI, you’ll immediately stand out. Modern AI recruiting tools are good at picking up nuance (I’m a fan of Juicebox 🧃). But they can’t spot what’s nonexistent. If evidence of AI is nowhere to be found in your profile, you won’t show when a hiring manager plugs in “AI-native” as a job requirement. Ofc, when you do, please make sure it’s real. And, no, chatting with an LLM doesn’t count in the same way that using Google doesn’t make you an SEO expert. === P.S. Scrunch is hiring like crazy. And, you can probably guess what trait we’re screening for 🙂 Dropping the roles I’m hiring for below 👇
-
Kevin White shared thisI can't tell if AI + software is making me hyper-intelligent or slowing removing the need to think. Case in point: The evolution of how I write (essentially, the mechanism for clear thought). A brief personal history: 1983 — born (no strong opinions yet) 1990s → ~1999 — Writing by pencil & paper ~1999 → 2024 — Writing by typing 2025 → present — Writing by dictating 2035 → ??? — Writing by just thinking?! ===== At each level of abstraction, I feel like bits of critical thinking are slowly slipping away. But at the same time, I feel like I'm 10x more productive. Which is it? Maybe both are true 🤷 #existentialfriday #onshottedbyai #tgif
-
Kevin White shared this🆕 Feature drop: Content Gaps tl;dr: Scrunch now tells you where and why you’re not showing up in AI answers. How it works: ✦ Scrunch crawls your site and cross-references it with the prompts and citations your brand is (or isn’t) showing up for, but should. ✦ From there, we auto-diagnose and surface content gaps (hence the name Content Gaps) and where to prioritize. Simple as that. Content Gaps is now live for everyone using Scrunch (7-day free trial if you’re not already) Go deeper 👉 https://lnkd.in/g7VSQQ7i 🎬 See it in action 👇
-
Kevin White posted thisLinkedIn these days 👇 Perception: “I replaced my team with 12 agents in a weekend.” Reality: “Can you help me rewrite this email?” In almost every conversation I have with a peer or candidate, I ask how they’re using AI. And, despite the LinkedIn rhetoric, most are still very early. That gap is your opportunity. But it’s closing fast. (Fwiw, I think we have ~12–24 month window for things to shake out) Those who bridge the gap to AI-native will be extremely hireable. Those who don’t will be… idk?, not so well off. So what does it actually mean to “bridge the gap”? 🪜 I put together a simple 4-level ladder (a maturity curve, if you will) to map where you are and which rungs to climb up: Level 1️⃣ AI assists AI helps you think. Examples: ✦ Rewrite this copy for me ✦ Find this stat to prove my point ✦ Summarize this document for me Basically, this is just ChatGPT as a sounding board. (The truth is that this is where most of the market is at today.) –––––– Level 2️⃣ AI builds AI helps you create. Examples: ✦ Generate a landing page ✦ Vibe code this dashboard ✦ Turn this document into a presentation At this stage, you’re producing shareable artifacts (not just internal slop) that are used for an external offer. Author’s note: I talk about AI a lot on here. I’ve been labeled as an “expert” on AI. Even then, I’d grade myself as a 2-2.5 on this spectrum. I say that not as a humble brag, but to show just how closeable the gap is if you jump in now. –––––– Level 3️⃣ AI systems AI makes you work repeatable. Examples: ✦ Generate on-brand visuals using a shared design system ✦ Turn call transcripts into messaging → campaigns ✦ Convert a long-form video into a fully content package At this stage, you’re not just prompting but building systems that give yourself (and your team) more leverage—i.e., you’re scaling knowledge. –––––– Level 4️⃣ AI agents AI does the work autonomously. Examples: ✦ Monitor product updates → automatically update docs ✦ Listen for relevant Reddit threads → draft a thoughtful response ✦ Track competitor launches → update battle cards This stage is pretty close to having AI agents as employees. Sure, those agents are doing the menial work that no one wants to do (that’s ideal in fact), but you’re reducing headcount or contract work at this stage as agents are actually replacing that work. ====== To close things out, my takeaway is don’t be discouraged by the “I’ve automated my entire workforce” crowd. But do take it to heart. It might feel like the ground is moving under your feet, but it’s not too late, and you’re not far behind. No better time to dive in than now.
-
Kevin White reposted thisKevin White reposted thisNew ship from the Scrunch team: Citation segments 🏷️ 3rd-party sources shape AI answers. But “3rd-party” doesn’t tell you a whole lot about them. Is it social media? A news source? An industry publication? A review site? Citation segments will tell you. Here’s how: ✦ Create custom citation segments in Scrunch. ✦ Name them, color-code them, and decide which domains match. ✦ See segments in your citations widget and citations table. The result? Instantly see who AI is citing for the prompts you care about, segmented by the sources that matter to your team. See it in action in the video below 👇
Kevin White commented on a post
2d
Not to step on your point but here’s a post on this and I work at a GEO tool company: https://www.linkedin.com/posts/kevbosaurus_microsoft-just-opened-up-pandoras-box-giving-activity-7429568791602409472-_ibZ
Fwiw, I can’t wait for the LLMs to start opening up the transparency around volumes. It’s guaranteed to come with ads. Advertisers won’t put up with ambiguity.
Kevin White commented on a post
2d
Shameless plug for a tool we build at Scrunch to address prompt tracking bloat. Essentially it tells you what prompts being tracked are redundant to cut and optimize cost.
Walkthrough: https://youtu.be/L2pkC4HXyTQ
Kevin White replied to a comment
3d
Ayush Poddar it's a good point. I do have the privilege of running with this stuff without many handcuffs.
-
Kevin White liked thisKevin White liked thisSometimes life gives you immense opportunities when you least expect them. BIG moves happening for me this Monday and I couldn’t be more excited. I’m stepping away from SDR Leadership, but moving in an adjacent direction that ignites a deep fire in me. With a leadership team I’ve looked up to throughout my whole career. Stay tuned 👀
-
Kevin White liked thisKevin White liked thisWe just launched Behind the Build, a podcast about the real conversations happening in construction. Not the project recaps. The calls you second-guessed. The risks that kept you up. What it actually cost to build something. Episode one is with Ryan Wall, CEO of SVM (Silicon Valley Mechanical). He and his partners started with almost nothing, took no salary for two years, and grew the company from $20M to $600M in revenue. One of the parts that stuck with me was Ryan talking about putting everything on the line early on. Personal finances, family support, all of it. Just to get the company off the ground. He talks about leveraging his house and borrowing money from every aunt and uncle he'd ever met. Not because things were going badly. Because he was all in and knew it. That's the kind of perspective we want more of out in the open. Clearstory CEO Cameron Page is hosting. The format is simple: sit down with GCs, subs, and industry leaders and get past the polished version of their story. If you're in the industry, this will feel familiar in a good way. If you're not, it's a pretty eye-opening look at what really goes into building anything. Episode one is live. Link in the comments.
-
Kevin White liked thisKevin White liked thisRelevance > Personalization. Relevance at scale is what AI agents can now unlock. I’m not talking scanning a LinkedIn profile and weaving in that somebody went to a school or likes golf. Custom Research/Signals let's you: 1. Scan your target accounts and find all check outs that have shipping prices set via order value 2. All e-Commerce stores that currently are offering discounts 3. All SaaS companies that have had outages in the last month 4. All relevant GitHub repositories for an open source company This used to be manual. What do you think of the email? #outbound #sales
-
Kevin White liked thisKevin White liked thisClarify launched "Prospecting" today! You can build a list + launch a campaign without leaving Clarify. Here's how you can launch your first campaign in under 3 minutes (it's magic): Step 1: Use natural language to create a list of people (data is automatically enriched). Step 2: Add those people to an email campaign (it uses AI to automatically create copy for you). Step 3: Review & Launch! It's so simple, yet powerful, that it feels like magic when you build your first campaign. [I'm grateful to Chris Eberhardt and the Clarify team for backing The Signal (.club) early on and continuing to support our work, as we help founders and revenue leaders bring AI into their GTM motion. They're building at the bleeding edge of AI and GTM.] If you're a start-up founder (or go-to-market leader), I highly recommend checking out Clarify → "The autonomous CRM that updates fields, manages pipeline, and answers questions—so you can close more deals." I recorded a video showing an example of a campaign for revenue leaders here in Austin, Texas, that I want to invite to an AI x GTM event I'm doing next month. Check it out below, and let me know what you think! This is what the CRM of the future looks like: AI-native, autonomous, and all-in-one. Congrats on the launch of Prospecting today! Patrick Thompson and the Clarify team are cooking! 🫡
-
Kevin White liked thisKevin White liked thisIf your brand is invisible to AI, it’s invisible to customers. Here’s where to start 👇 Scrunch cofounder and CEO Chris Andrew is moderating a session for Association of National Advertisers members featuring John Lovett, VP of Analytics at Seer Interactive. The agenda? Practical strategies for improving brand visibility in AI search (aka things you can start doing today). Count on: ✦ Actionable measurement frameworks ✦ Real-world examples from market leaders ✦ Concrete next steps for shaping how AI interprets your brand 📅 April 8 | 1-130 p.m. ET This webinar is ANA member-exclusive. If that means you, register here: https://lnkd.in/gEUpQgaR
-
Kevin White liked thisKevin White liked thisEarlier this week, I wrote about how your brand can show up in AI answers without ever being mentioned. The obvious follow-up: why do AI answers rely so heavily on third-party sources? Because AI doesn't default to your website. It's not visiting your homepage, forming an opinion, and deciding you're credible. It's building an answer from what it can reference across the web - editorial coverage, reviews, forums, directories, industry publications. Your site is just one input. Rarely the main one. The data backs it up: Scrunch Data Labs (Dec 2025–Mar 2026): ~87% of citations in unbranded prompts (prompts that don't include the brand name in the query) come from third-party sources At the account level: 5,547 LLM responses across 215 non-branded prompts → 84% third-party External analysis (23K+ citations): owned content makes up ~23% on average And the largest sample yet - across ~500M citations in Scrunch Search data, 89.6% of unbranded prompts are fulfilled by third-party sources. That number barely moves by platform (89.0% to 90.1%). The behavior is consistent regardless of which AI you're asking. Same pattern everywhere: models lean on third-party sources as neutral proof. A brand can say anything on its own site. A journalist, review, roundup, or community thread is harder to manufacture. That external layer is what creates credibility. What I'm seeing with clients: the brands owning AI visibility aren't just strong on-site - they're everywhere else too. Content strategy and PR are no longer separate. Control your narrative, but earn it across the ecosystem.
Experience & Education
-
Scrunch AI
View Kevin’s full experience
See their title, tenure and more.
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Recommendations received
2 people have recommended Kevin
Join now to viewOther similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top contentOthers named Kevin White in United States
2092 others named Kevin White in United States are on LinkedIn
See others named Kevin White