HubSpot lost 140m organic visits in a year after Google's 2024 algorithm changes and still grew revenue and demand.
Its first AI win came from auto prospecting emails built from top sellers' language, delivering 10x conversion gains after six months.
HubSpot now leans on a 350-creator media network that generates more demand than its search-driven content did.
HubSpot rebuilt marketing for AI after losing 140m search visits
HubSpot built its growth on search-driven content, and AI erased that channel, costing it 140m visits in a year. Flanagan's answer is to rebuild marketing around AI agents trained on HubSpot's own context and to replace search with a creator network that earns citations in answer engines. The book Loop sets out the method, with an email experiment that lifted conversions tenfold as the early evidence.
Takeaways
Modular AI email prompts produced 10x conversions. HubSpot wrote the prompts with its sales team, in the language of top sellers and separate blocks for subject line, hook, value offer and call to action. The test took six months to reach 3x, 5x and 10x conversion gains, breaking Flanagan's six-week rule.
HubSpot lost 140m visits and kept growing. After Google's 2024 algorithm changes HubSpot lost 80% of its organic traffic, 140m visits in a single year, and still grew revenue and demand. Flanagan reads that as proof the old search-led model could not hold.
A creator network replaced search distribution. HubSpot bought The Hustle in 2021 in a deal Flanagan led, then built a network of 350 creators across four tiers on two-year contracts, taking half of their ad inventory. The network now generates more demand than search content did.
Hiring great people builds the AI context layer. Hiring great people enriches a context layer that trains a company's AI agents, so their intelligence stays with the company after they leave. Flanagan argues context, not model access, is the durable advantage in the AI age.
AI creation delivers the average of everyone's taste. Asking Claude or ChatGPT to create things gives you the average of everyone else's taste, and AI's polish now reads as generic slop because it is everywhere. Junior marketers struggle most because they cannot yet teach AI what good looks like.
Informational search is dead, transactions survive. AI killed informational content while transactional search still works. 80% of B2B buyer journeys used to start on search and 80% are predicted to start in answer engines by 2027, so visibility comes from citations rather than clicks.
HubSpot runs many big bets at once. The Claude connector is an exponentially growing source of users, and an ads integration with OpenAI was a first for a CRM. Product-led growth since 2015 means over half of HubSpot's revenue now comes from free users.
Loop's four stages
Express. Taste and judgment decide what gets made. Making things was the hardest part of marketing for a hundred years, and AI removed that, so the question became what to make.
Tailor. Work is adapted to individuals rather than segments. Taste profiles carry rules per persona, since HubSpot sells to different buyer personas.
Amplify. A new set of distribution playbooks for marketers.
Evolve. How to restructure a marketing team for the AI age.
Tap a name to go there · tap a line to see why they are linked