Loop marketing, from HubSpot's new book: what it confirms about how a website should earn

HubSpot spent fifteen years teaching the world inbound marketing: write blog posts, rank on Google, collect leads. Its new book says that playbook is over. Loop: Outlearn. Outmarket. Outgrow., by Kipp Bodnar and Kieran Flanagan, published on 22 September 2026, argues that the teams that win now are the ones that learn fastest, not the ones that publish or spend the most.¹

The headline story: HubSpot's blog lost about 80% of its traffic, and revenue kept growing, because most of the lost visits were people who were never going to buy.² If the company that invented inbound measures revenue per visit instead of visits, every business with a website should look at what it measures.

We set the book's main ideas against the way we run websites, for clients and for ourselves. This piece works from the publisher's page and two long interviews with Bodnar; where a figure is the authors' own claim, we say so.

The loop, in one paragraph

The book describes four stages that repeat: Express (say clearly who you are and what you think), Tailor (fit the message to who is reading), Amplify (take it where buyers already spend time) and Evolve (read what happened and change the next round). Each round is scored on four measures it calls the four Rs: Remarkability, Resonance, Reach and Regeneration, the last one being how much better each cycle gets than the one before.¹ ²

1. Measure what a visit is worth, not how many there are

The book: track revenue per visit and the pipeline content brings, not traffic. Informational posts that never sold anything were most of what HubSpot lost, and it barely mattered.²

How we work: this is where our audits start. The free audit looks at what stops a visitor from becoming a lead. The deep audit reads the client's own Search Console and analytics and ranks every problem by what it is worth, because an audit should end in money, not in a checklist. We sort searches by intent and set aside the ones that look for a fact and will never click through to buy (which searches are worth money). And every lead on the sites we build arrives with its source, so a page is judged by the enquiries it brings.

2. Keep, fix or remove each page by what it does for the business

The book: HubSpot pruned thousands of old blog pages and rewrote or merged the ones that produce business.²

How we work: on one site we cut most of the pages and traffic went up (less site, more traffic). The deep audit lists the pages with no click in twelve months and, just as important, the quiet pages that still earn, so nothing that pays goes to the bin by mistake. When we rebuild a site, every old address is classified keep, forward or remove before launch.

Where we differ: we decide by search demand as well as past results. A page that never earned may sit on a search people make every day; the fix is a better page, not a deletion.

3. Define the voice before the AI writes

The book: build a "Taste Profile" (who the buyer is, the brand's tone and point of view) and give it to the AI with every task, plus a list of banned words. Bodnar grades generic AI output a C and Taste-Profile output an A; those grades are his own.³

How we work: it is the first deliverable of our growth engine: one page in plain words on who you sell to, what they ask and how you sound, which every agent works from. Next to it, a facts file with a source for each claim and a list of what is never said. We write the rulebook before the first page, and a person checks what ships.

4. Build on what only you have

The book: AI makes generic content free, so generic content is worth nothing. Build around your own data, customer stories and results.³

How we work: our rule for AI content is to write only what you can source: every fact comes from a record, a page or the client's own data, and anything we can't source is held rather than guessed. The same idea runs through the AI systems we build: they work on the client's own data and method, not on a generic model's guess.

5. Learn in short cycles

The book: Regeneration is what makes the loop compound. The authors claim a company running 26 learning cycles a year will outperform one running four; that is their argument, not a measured result.²

How we work: our ongoing engagement runs in monthly cycles: each change is checked against the client's own numbers, what worked gets more, and what didn't is reported as such. We decide what success means before a test starts, so a result can't be read to fit.

Where we differ: speed is right where a change is cheap to undo: a title, a call to action, the timing of a post. It is wrong where a mistake costs rankings: addresses, redirects, a site's structure. There we test before launch and compare every page (redesign without losing what ranks). Fast where it's safe, careful where it isn't.

6. Answer the questions people now ask AI

The book: people type 2 to 4 words into Google and 40 to 100 into an AI assistant. Write plain answers to full questions, and track separately whether AI answers name you and whether they link to you.²

How we work: our free audit already flags questions on a page whose answer is buried under a long paragraph, because AI assistants and search results quote the first lines under a question. Tracking mentions in AI answers is newer ground for everyone. It needs no tool to start: each month, ask 25 full questions your buyers would ask in ChatGPT, Google's AI Mode and Claude, and write down whether each answer names you, links to you, or neither.

7. Own the audience, don't rent it

The book: about 90% of HubSpot's content leads now come from newsletters, creators, podcasts and video, and about 10% from the blog.²

How we work: the growth engine runs the client's social channels and blog as one routine, each post leading back to a page with a next step (the weekly routine). Free tools that give the answer before they ask turn readers into contacts you can write to. A list you own is the one channel no algorithm can take away.

What any team can start this month

• Pick one number per visit: leads or sales per 1,000 visits, by page. Read it monthly, against the month before.
• Write the voice page before the next piece of AI content: buyer, tone, point of view, banned words.
• List your pages with no enquiry in a year, then check each against search demand before deciding.
• Run the 25-question AI check once a month, in a spreadsheet.
• When a test's result is due, write the next test the same day, so the loop never stops at a report.

If you'd rather see first where your own site loses visitors before they become leads, the free website audit shows it from the outside, in a few minutes.

Notes. ¹ HubSpot, “Loop” book page (publisher's description). ² Greg Jarboe, Search Engine Journal, “HubSpot Lost 80% Of Its Traffic: CMO Kipp Bodnar Explains Why Revenue Kept Growing”, 29 October 2026 (an interview; the article itself flags some figures as unverified). ³ Duct Tape Marketing, “How Inbound Marketing Is Changing in the Age of AI”, podcast with Kipp Bodnar. Read on 30 October 2026.