The new Helpful Content, and the 7‑step framework to win AI search.
What Google told the room in Toronto, translated into a playbook you can run on Monday, with worked examples across finance, travel, edtech and e‑commerce.
In April 2026 at Search Central Live Toronto, Danny Sullivan gave the industry its clearest content distinction yet, and quietly retired the “helpful content” label in the process.
Google’s framing is simple: there is commodity content and non‑commodity content. It reinforces what John Mueller said in May 2025. Focus on making “unique, non‑commodity content.” Good non‑commodity content rests on three pillars:
Brings a viewpoint, dataset or insight others lack and can’t easily replicate.
Talks about one real situation, not general rules, steps or roundup advice.
Demonstrates first‑hand knowledge or experience: not rewritten search results.
Sullivan’s reassurance: you don’t need to rip your site apart, fragment everything for machines, or chase every query variation. The fundamentals still win, AI search rewards the same quality traditional search always did.
Here’s the test we apply to every page we ship: if a model can answer the query from its training data alone, your content is commodity, and it will never cite you. If the answer can only come from you, you become a required citation. Non‑commodity content is simply the discipline of manufacturing that requirement on purpose.
Commodity content isn’t banned, it’s a natural part of content strategy. It’s just a weak play for clicks or AI citations in 2026.
Generic, replicable, rewritten from what’s already on the web.
“Top 10 Things to Consider…”, “What is X?”, surface-level overviews, generic tips.
Anyone with a content brief and an internet connection can produce it. AI can reproduce it for free, so it doesn’t need to send a click, or a citation, your way.
First‑hand, specific, proprietary, content only you could write.
Real client work, original data, a named situation, a hands‑on test, a contrarian POV.
The kind of material AI must surface, because it’s the only place the information exists. Harder to replace, far more likely to be referenced in AI answers.
Note: “commodity content” isn’t an officially confirmed ranking term, it’s Google’s framing for the content that no longer earns visibility in the AI era. Useful as a strategy lens, not a settings checkbox.
A model can state “personal loans need a good credit score” with no source, it’s common knowledge. It cannot state “approval rates fell 9% after the March bureau-policy change” without naming where that number came from. The more your page is built on facts that originate with you, numbers, tests, named cases. The harder it is for an AI to answer without you. That pull is citation gravity, and it’s the mechanism that turns Google’s non‑commodity push directly into AEO/GEO wins.
The same line, commodity vs non‑commodity, drawn through the industries we work in every day.
“5 Tips to Improve Your Credit Score.” · “What is a Personal Loan?”
“We analysed 12,000 rejected loan applications: the 3 fields that actually decide approval in India.”
Why it wins in AI: first‑party data a model can’t reproduce, so it becomes the cited source for “what affects loan approval,” not a forgettable explainer.
In finance you often can’t publish raw client stories, compliance won’t allow it. So non‑commodity shifts from war stories to proprietary frameworks, anonymised first‑party data, and expert interpretation of regulation. In regulated markets, the moat is the analysis: not the anecdote.
“Top 10 Things to Do in Goa.” · “Best Time to Visit Bali.”
“We ran Mumbai–Goa by bus, train and flight for 30 days: real costs, real delays, the hours nobody quotes you.”
Why it wins in AI: hyper‑specific, seasonal, first‑hand operational detail that aggregators and AI simply can’t fabricate from training data.
Two verticals where the cited answer almost always belongs to whoever holds the real data.
“Benefits of an MBA.” · “Top 5 Online Courses for Data Science.”
“We tracked 3,400 alumni for 5 years: the real salary curve by specialisation: not the brochure number.”
Why it wins in AI: outcome data and cohort insight only the institution holds. It becomes the cited answer to “is this program actually worth it?”
“2026 Kitchen Trends You Need to See.” · Manufacturer‑copy product descriptions.
“We stress‑tested 14 ‘premium’ running shoes to 400 miles, the exact mile each one failed.”
Why it wins in AI: teardown and real failure‑mode content. It’s the source a model reaches for when someone asks “which one actually lasts?”
So its marginal SEO value is collapsing toward zero too. The only durable asset left is content with an effort or experience moat a model can’t cheaply reproduce, which is exactly the signal Google surfaced as contentEffort in its 2024 documentation leak: an estimate of the real human work behind a page. High‑volume, low‑cost content isn’t risky because it’s AI‑written; it’s risky because it’s effortless.
We consolidated our nine‑stage model into seven and rebuilt the content engine around non‑commodity content. Each step builds on the last.
Read it this way: Steps 1, 2 and 7 are the SEO fundamentals Google reaffirmed in Toronto. They remain foundational. Step 3 is where the new battle is actually won.
Straight from the Toronto stage. The over‑engineering you can safely drop.
“Conversational keywords.” No need to chase every synonym or query variation, Google’s language matching already maps your page to many queries.
“Chunk it for AI.” Don’t fragment content for machines. Organise and write for a good human reading experience; use H1/H2 for readers, not machine‑precision.
“JavaScript breaks AI.” JavaScript is fine, as long as Google can access it the way a human does.
“AI content is the problem.” Using AI for research and structure is fine. Using it to mass‑produce pages without adding value is what trips Google’s scaled‑content‑abuse policy.
| Traditional Search | In the AI era |
|---|---|
| Content | Prioritise non‑commodity content |
| Page experience | Stays foundational |
| SEO fundamentals | Audit for gaps |
| Structured data | Expand, especially in e‑commerce |
| Shopping / Local / Video / Image SEO | Review for new opportunities |
| Agentic search | Watch & experiment early |
The content‑creation floor collapsed, so Google lifted the quality ceiling for what gets indexed at all. “Crawled. Currently not indexed” is now usually a quality verdict, not a bug. If pages aren’t getting in, the fix isn’t more posts. It’s better source material.
We don’t ask AI to invent expertise. We extract it. Operator interviews, real client data, hands‑on tests and original research become the raw material; Butterswipe handles prompt tracking, structure and competitor‑led authority mapping. We find the prompts you must own, then build the first‑party assets that make you the required citation.
Sources: Google Search Central Live Toronto, Danny Sullivan, Director of Google Search (April 2026); Google Search Central Blog, John Mueller (May 2025); Google Search Essentials & scaled content abuse policy. Original analysis, frameworks and vertical examples by 6sMarketers.
