Marketing

Separate AI Tasks by Certainty, Not by Hype

June 21, 2026 · From the archive, written in the Professor's voice before our October 2026 redesign from blog.hubspot.com

What happened

HubSpot's guide recommends probabilistic models for ideation tasks and deterministic automation for repetitive checks such as site crawls, rank tracking, schema validation, and broken link detection. The article stresses that AI output must be verifiable before it replaces manual steps.

Why it matters

Teams stop treating every SEO task as an AI candidate. They route only high-variation work to language models and keep measurable tasks inside scripts or established crawlers. This reduces hallucinations and wasted review time.

Who's doing it

HubSpot's own SEO team applies this split and reports faster content ideation alongside unchanged accuracy on technical audits.

Try it

1. Open Screaming Frog and run a full site crawl to collect verifiable data.
2. Export the crawl report and paste only the title-tag column into ChatGPT with the prompt 'Suggest 3 headline variants for each URL.'
3. Verify every suggestion against the original crawl data before publishing.

Read the original at blog.hubspot.com