Scrunch Spent Two Years Teaching Robots to Read Your Website - Then Sold for $225 Million

For twenty years, marketers played a single game: get to the top of Google. You wrote the keywords, you earned the links, you waited for a blue link to move up a list. Then people stopped scrolling lists. They started typing questions into ChatGPT, Perplexity and Gemini and reading the paragraph that came back. Scrunch was built for that exact moment - a Utah software company that shows brands how they look inside an AI answer, and then helps them look better.

The company started in 2023, well before "AI search" had a budget line at most companies. Co-founders Chris Andrew and Robert MacCloy had both been early employees at Hearsay Systems, a compliance-heavy software firm that sold to large, cautious enterprises. That background matters: they did not build another consumer chatbot. They built the measurement layer underneath the chatbots, aimed at the kind of marketing and brand teams they already knew how to reach.

The pitch Andrew repeats in interviews is blunt, and a little frightening if you run a website for a living.

Your website doesn't need to go away, but 90% of its human traffic will. AI agents are becoming the new front door to your brand. Chris Andrew, CEO and co-founder

What it actually doesTurning an AI's opinion of you into a dashboard

The core product answers a question every brand suddenly has and cannot see on its own: when someone asks an AI about my category, what does it say, and do I show up? Scrunch takes the prompts a real customer might type - "best noise-cancelling headphones for travel," "which payroll software works for a 40-person startup" - and runs them across the major engines on a recurring schedule. It records whether the brand appears, in what position, with what sentiment, and which sources the AI cited to get there.

That produces the metric at the center of the whole category: share of voice inside AI answers, tracked against competitors over time. From there the platform flags the gaps - a product the model never mentions, a fact it gets wrong, a competitor it favors - and points to the content that would fix it. A separate feature watches the server logs to show which AI crawlers are actually visiting the site and which pages they read, and ties that back to real referral traffic through a Google Analytics integration.

The differentiatorTwo versions of your website - one for people, one for bots

Monitoring is now a crowded room. The more distinctive piece is the Agent Experience Platform, or AXP, which shipped alongside the 2025 Series A. Instead of only reporting on how AI reads your site, it sits at the delivery layer and serves AI agents a compressed, structured, machine-readable version of your content - without changing anything a human visitor sees. In practice that means one site for people and a cleaner, parallel feed for the machines doing research on their behalf.

They're not just observing the AI shift. They're helping brands rewrite the web for AI agents. Jon Sakoda, Decibel

Andrew's second argument is why marketing teams stopped treating this as a science project. A visitor who arrives from an AI answer, he says, is buying faster than a traditional search visitor, because the bot already did the comparison shopping. Read that way, AI referrals are not lost clicks - they are pre-qualified buyers arriving with the homework done. That reframing turned a defensive panic ("AI is stealing our traffic") into an offensive budget line ("win the AI answer").

GEO vs SEOThe same anxiety, a new algorithm

The category has a name now - Generative Engine Optimization, or GEO - and it rhymes with the discipline it is quietly replacing. The mechanics, though, are different enough to need new tools.

Rivals have crowded in - Profound, Peec AI, Otterly, Athena, Bluefish and legacy SEO suites bolting on AI features. Scrunch's edge in the race was less about having every feature and more about pairing measurement with that delivery-layer product, and closing enterprise logos while the category was still being named.

The expertise behind that came from an unglamorous place. Both founders spent roughly a decade at Hearsay Systems selling to banks and insurers, industries where "how a brand is represented" is a legal question, not a marketing one. That taught them to build for teams that need audit trails, alerts on misinformation and proof they can hand to a compliance officer. It is the same instinct now aimed at AI answers: when a machine gets a fact about your product wrong, Scrunch treats it as a defect to log and fix, not a talking point.

Who is payingFrom 25 customers to a Sitecore-sized problem

Early customers were the kind of names that make a Series A deck work: Lenovo, Penn State University, the software developer BairesDev. Within about a year of the public launch the roster passed 500 brands and agencies, adding Skims, Headspace, Akamai, Clerk and Runpod. These are teams with real reasons to care what a machine says about them - a hardware giant that does not want to be misdescribed, a wellness app that lives or dies on trust, an agency that needs to prove AI visibility to its own clients.

The business model is unfussy B2B SaaS: no free tier, a short trial, then subscriptions that scale with prompts, workspaces and seats. Brand plans start near $250 a month and rise to custom enterprise pricing; agencies start around $500 with dedicated workspaces for pitching new clients. It is priced like a tool a marketing team keeps, not a one-off audit.

The exitWhy a legacy giant paid nine figures for a two-year-old

Scrunch raised about $26 million in total: a $4 million seed led by Mayfield in March 2025, with angels including Hearsay co-founder Clara Shih, PillPack's TJ Parker and Webflow's Bryant Chou, then a $15 million Series A led by Decibel in July 2025. Roughly a year after that, in June 2026, Sitecore - a long-established digital-experience company - acquired Scrunch for a reported $225 million. Andrew stayed on to run it inside Sitecore.

The logic of the deal is the most interesting part. Sitecore's entire business is helping companies manage their websites. Paying $225 million for a startup whose core belief is that the human website is becoming a data feed for machines is, in effect, an incumbent buying its own future before the future arrives. When slow, well-capitalized companies get scared enough to write that check, it tells you where the market thinks it is going.

Your program is no longer what you say it is. It's what you say it is, plus third parties, plus competitive sites. Chris Andrew, on losing control of the brand narrative

What you can copyThe playbook underneath the headline

Strip away the acquisition and a few reusable moves remain. First, the founders sold to a customer they already knew - their compliance-software past told them exactly which enterprise buyers to call. Second, they named a fear the whole industry could suddenly feel and sold the aspirin for it, rather than waiting for the category to mature. Third, they did not stop at a dashboard; the delivery-layer product is what turned a reporting tool into something an incumbent wanted to own. And fourth, speed of narrative beat depth of revenue - a credible claim on an inevitable shift, closed while competitors were still arguing about definitions.

The obvious caveat: this worked because a platform shift was genuinely underway and the timing was early. Run the same play into a settled market, or a year too late, and the "measurement layer for the next big thing" is just another tool nobody has budget for yet. Scrunch's bet was that AI answers would become the front door before most people were ready. It got about two and a half years of runway to be proven right, and a reported $225 million says it was.