Search Disrupted Newsletter (Issue 16)

Google’s AI Mode for everybody

Google’s IO conference this week consisted of them ritualistically chanting “AI! AI! AI!” on stage while making product announcements.

Sundar Pichai and crew announced a ton of new search and search-adjacent products, and while there are a lot of headlines about the exact features announced, etc.

The main point is being missed: Google is telling you that Gemini is now the new centerpiece of the search ecosystem.

Note: this doesn’t mean that people are going to stop using Google search, but that if they “Google” something it’s going to be asking Gemini, they’re going to get a qualified editorial answer from Google (generated by Gemini), and maybe some links if it makes sense.

Currently, we’re seeing this introduced as the “AI Mode” in Google Search, which does a “pre-search” of sites before displaying a shortened set of results.

AI Mode is graduating from Google Labs and is now available to everybody. This follows the same rollout strategy that AI Overviews took.

Link: https://search.google/ways-to-search/ai-mode/

John Doherty on AI Mode

For a more in-depth examination of AI Mode, my friend John Doherty launched a new YouTube channel for his company, Editor Ninja.

He covers the implications for SEO and has some solid suggestions for your next steps.

Link: https://www.youtube.com/@editorninja

Knowatoa Knowledgebase launches

As I personally learned this week, it’s hard to write up knowledgebase articles when you have to keep defining new words to capture the changes happening in AI search. But we did it!

The new Knowatoa Knowledgebase is life at the link below for both Knowatoa users as well as for any SEO who wants to learn the differences between “Search Volume,” “AI Chat Search Volume,” and “AI Search Volume.”

We’ve also started adding in-app chat support that ties in with the knowledgebase and the existing helpdesk. You can email us at support@knowatoa.com - we’re grateful for the steady stream of folks sending us feedback, suggestions, and bug reports.

So, please check out the knowledgebase, and if something is missing you’d like to see, please let us know.

Link: https://docs.knowatoa.com/

How LLMs Fail at Decision Making

It’s frustrating and unintuitive that LLMs act more like people and less like computers.

Case in point, a new study by the Collective Intelligence Project on how LLM models make judgments about what’s “best” in any given situation is significantly influenced by things like:

• what order the options are presented
• labels on the scale (1-5 stars and A-F) while having the same meaning get scored differently
• the way the options are described

Admittedly, we shouldn’t be surprised that Large Language Models are influenced by language, and while one take on this is that AI models are “bad” and should never be used, my take is that we need to be on guard for biases and issues the same way as human judgments.

At Knowatoa, we track these types of quirks in the different models as they often point to areas where our clients could be misrepresented online or where there’s an opportunity to rank higher or be recommended more strongly.

Read the full study at: https://www.cip.org/blog/llm-judges-are-unreliable

Knowatoa Updates

Check out our updates page for details on how we’ve:

1. Improved the refresh times of tracked questions
2. Added new filter options for sorting through the list of suggested questions to find ranking opportunities

Link: https://knowatoa.com/updates/

Thanks

Thanks to Lily Ray (fresh off DJ’ing at SEOweek) for her shout-out to Knowatoa on stage at the AWIN conference in Portugal. Photo credit to our friend Monica Lent, who was there to capture the moment.

p.s. It would really help me out if you could Follow me on LinkedIn