1) 1000s of Google Search API documents leaked

This has been a big ‘oops’ for Google, as SEO’s are discovering that prominent Googlers have been misdirecting them about how their ranking systems work.

According to Rand Fishkin, “On Sunday, May 5th, I received an email from a person claiming to have access to a massive leak of API documentation from inside Google’s Search division.”

The source asked to remain anonymous, but on a video call went on to disclose other private information to Rand, who shared what he discovered in an article published on May 27th.

Here are some influencing features we now know to exist that have previously been denied by Google spokespeople.

1. Google do use click-based user signals within their ranking systems
2. There is a sandbox to segregate websites based on trust signals, one exists “to sandbox fresh spam in serving time”
3. They do use Chrome data to influence organic search
4. They do use domain authority. Documentation refers to a feature called ‘siteAuthority’

Mike King has written a detailed exploration on these documents, which you can find here.

2) Consent mode v2 changes cookie tracking

Google has recently updated their consent mode policies to conform with the Digital Markets Act (DMA).

This is EU legislation, but they are making their policy applicable to the UK. This change supports a shift to cookieless tracking.

Consent Mode v2 essentially adds two additional flags that signal:

1. Does the user consent to their personal data being used for advertising purposes? (ad_user_data)
2. Does the user consent to their personal data being used for remarketing? (ad_personalization)

Websites that fail to comply may have their ads accounts suspended by Google.

How we’re ensuring clients are compliant

We’ve been implementing new cookie banners that allow users to give explicit consent to being tracked for advertising or re-marketing purposes.

If users don’t interact with the banners, or decline being tracked, they will not be tracked. Over time, Google Analytics will model the data of users who aren’t being tracked. This ensures their privacy is protected, and that the clients sites comply with Google’s policy.

Business owners: if you’re unsure whether your site is Consent Mode v2 compliant, get in touch with our digital team.

Read more about staying ahead in a cookieless future.

3) VideoFX - Google’s new AI video generation application

VideoFX is Google’s latest development into generative video AI, powered by Google DeepMinds Veo. This is Google’s newest video generation model, and is its most capable advancement in the software so far. It is packed with features, but notably, the tool includes a storyboard mode and advanced cinematic effects.

They are also releasing sought after updates for ImageFX, including:

• Editing controls that will allow users to simply brush over elements in an image to add, remove, or change them.
• Imagen 3, which is Google’s most advanced, capable image generation model. It benefits from more accurate text rendering, and greeter photorealism.

And updates for MusicFX, like:

• DJ mode, a feature that lets users mix beats combining instruments, genres, and much more.

Unfortunately, the tools are only available in the US at the moment - but you can sign up to a waitlist to hear when they are available in the UK.

4) Google start to enforce their Site Reputation Abuse policy

Example of search changes, source: @glenngabe on X

Google reviewers started to hand out penalties to sites who were breaching the policy on May 6th. However, this hasn’t yet been applied to the algorithm yet - so more penalties might be coming.

“We’ll now consider very low-value, third-party content produced primarily for ranking purposes and without close oversight of a website owner to be spam.” - The Keyword - Google’s Blog.

Examples of site reputation abuse include:

• A third party publishing payday loan reviews on a trusted educational website
• A third party hosting a coupon directory on a news site

The biggest impact has been seen amongst publishers.

5) Google I/O 2024 announcements

Google I/O is Google’s biggest developer conference. This year, they announced lots of exciting projects. Here are some highlights.

• Google AI Overviews - Google have evolved their Search Generative Experience into AI Overviews, and moved it out of Search Labs into live US search. More on this later on in this article.
• Project Astra - Google DeepMInds vision for the next generation of AI assistants. The idea is to create an assistant that is capable of having a natural conversation with quick responses.
• Gemini 1.5 Flash - This is a lighter, faster version of the new Gemini 1.5 Large Language Model, equipped for speed and multi-modal reasoning.

Check out all the announcements

6) AI Overviews go live in US search

Google SGE was moved out of its testing phase and into live US search under a new name, AI Overviews. Here are the key points:

When users type in a query, Google’s Gemini LLM (Large Language Model) generates an answer using search results as sources

• The idea behind this is to create a slicker, more helpful search experience - allowing users to interact with the AI Overview and let “Google do the Googling”
• AI Overviews are only being generated for a fraction of queries at the moment, and only in the US
• There is no way to turn AI Overviews off, which is causing complaints from searchers who don’t like the feature
  

It’s crucial to remember that this feature is experimental - any results should be treated with scepticism.

The impact on search

The full impact isn’t known yet since the feature hasn’t been completely rolled out, but here’s what we know so far:

• Less users clicking through to websites on queries that generate AI Overviews - more ‘zero-click search’.
• There will be less classic ‘blue links’ in search. Eventually, there will be a ‘Web’ filter showing just blue link results
• Alphabet/Google CEO Sundar Pichai said AI Overviews have resulted in an “increase in Search usage*.
• However, search experts argue that this means the AI Overviews aren’t doing their job - since the point of them is to let Google do the legwork. If people are searching more, this could mean their queries aren’t being answered by AI Overviews
  

There have been some questionable results

Results are using websites that don’t rank within the top 100 results on Google as sources.

This means results are being formed from low quality websites which don't follow best practices that have been encouraged by Google. In some cases, AI Overviews is getting information from AI generated spam articles. It is also a victim to AI hallucinations - sharing information and facts that don’t exist.

Because of this, and the nature of how LLMs work, it is far too easy to get the tool to generate dangerous, incorrect, and NSFW content for queries it probably shouldn’t be answering.

Results will improve with time

And the feature will, ultimately, make search better for users. However, it probably shouldn’t have moved out of testing so soon.

See you next time