Hey friend!

Last week, we reported on the release of o1 (or “Strawberry”) from OpenAI—a new AI reasoning model designed to spend more time thinking before it responds. Apparently that time thinking could be spent thinking about how to better deceive you? 

On the bright side, at least Google is taking steps toward AI transparency. More on both of these stories in this week’s news roundup! 

–Thenuka

[INSERT_SUBSCRIBE_CTA]

What’s new in AI and content?

🤥 o1 can lie

OpenAI’s newly launched AI reasoning model, o1, is off to an interesting start after the AI safety research firm Apollo Research found that it has the capacity to deceive. To be clear, this isn’t quite the same as AI hallucinations, which are more like inaccuracies resulting from misinformation, bias, and poor data quality. Instead, o1 shows instances of fabricating info and creating fake links when it can’t complete certain tasks. These deceptions are rare—about 0.8% of responses—but they’ve led OpenAI to rate o1 as a “medium risk” in persuasion and a terrifying category called “chemical, biological, radiological, and nuclear (CBRN) threats.”

• What this means for you: Take this as a reminder to fact-check any content you generate with AI. If you’re working with o1 specifically, know that the model’s “chain-of-thought" lets it simulate a thought process—so you can actually ask how it arrives at certain decisions. Doing so could help uncover weird gaps, even moments where o1 appears confident but is actually uncertain.  

👀 Google’s AI Overviews show more to signed-in users, especially for e-commerce searches

According to new research from Brightedge, Google’s AI Overviews (AIO) appear 10-20% less for signed-out users. This varies across industries, with e-commerce queries showing the biggest difference: 90% fewer AIO results for signed-out users. Brightedge’s report also includes new findings about AIO citations, the links that appear alongside the AI summaries. Looks like niche authoritative domains in healthcare, education, and tech get more citations than consumer sites and blogs. For example, in the healthcare space, arthritis.org saw a recent boost in AIO citations while the number of citations for verywellhealth.com declined. 

• What this means for you: 

  • From a consumer perspective, if you’re one of the many who dislike Google’s AI Overviews, try signing out of your Google account to see them less. 

  • Now from an SEO perspective, AIO’s new citation preferences can mean a change in your traffic depending on how trustworthy your domain is. To strengthen your site’s reputation, review Google’s E-E-A-T guidelines and make adjustments to your content accordingly. In his AIO analysis, SEO veteran Kevin Indig also emphasizes optimizing your content around user intent rather than keyword matching.

🖊️ Valid HTML and typos aren't ranking factors

Google Search’s John Mueller took to LinkedIn recently to clarify that while valid HTML and the lack of typos reflect content quality, they’re not ranking factors for Google Search. If they were, that’d set a “pretty low bar” for ranking. Still, having clean HTML and error-free content matters for user experience. Surprisingly, meeting this low bar might not be as common as you’d expect: according to one report, 99.5% of the 200 most popular websites feature some HTML that’s either invalid, nonexistent, or broken.

• What this means for you: Don’t skimp on good web design and writing/editing best practices. Even if it doesn’t directly impact rankings, having valid HTML and typo-free content enhances trust and makes for a better user experience all around.

🖼️ Google plans to make it easier to detect AI-generated images

Google is integrating tech from the Coalition for Content Provenance and Authenticity (C2PA) to help users more easily distinguish real images from AI-generated or edited ones. This is a big step in terms of content transparency in the age of AI, and will hopefully set a precedent for other platforms to follow. 

• What this means for you: Expect to see this change roll out to Google’s “About this image” feature in search results. It’ll tell you whether an image was taken with a camera, altered by software, or created with generative AI. Google’s ads will also integrate C2PA metadata, and there are plans for the same authentication tech to eventually be introduced to YouTube.

Playbooks & Guides

📝 How to replicate Canva and PandaDoc’s “template” pSEO strategy

What do Canva and PandaDoc have in common? They both use template-driven programmatic SEO strategies to drive traffic. In other words, they create content around template keywords like “birthday invitation templates” for Canva and "investment contract template" for PandaDoc. The results speak for themselves: more than 13 million monthly visits to Canva and 140,000 to PandaDoc. We explain who this template strategy works for and how to execute it in one of our latest playbooks.

>> Find out more

📈 Drive high-intent traffic with the “NxN” pSEO strategy

Zapier gets 280,000 monthly visitors with absolutely zero paid advertising. How? Aside from publishing robust blog posts, it creates app-to-app integration pages that rank for keywords like “Word to Docs” and “Excel to Google Calendar.” We call this the “NxN” programmatic SEO strategy—and Zapier’s not the only company making use of it. Check out our playbook to find out how the NxN strategy works and whether your site should give it a shot.

>> Find out more