Updates by Sr. SEO Brett Woodward
New rules under the EU’s landmark AI Act require companies to self identify products and content created with AI. As of August 2 new companies will need to comply before distributing their products and existing companies will have 4 months to comply.
This move largely looks to combat explicit disinformation such as the damage that can be caused by deepfakes, but anyone in the EU using AI in their products will be affected, marketers included.
The Guardian reports that, “Under the rules, synthetic text, images, video and audio designed to look truthful must be visibly marked as AI-generated and contain a digital watermark to show the artificial origins of the content.”
Anthropic has since committed to complying with these rules, and released a document outlining the inner workings of its watermarks that will now accompany all material that Claude generates, inside the EU and out.
As for the impact this will have on the field of search marketing, I’m seeing lots of chatter online about how Claude watermarking its outputs will make it much easier for Google to identify AI generated content and potentially make the outputs themselves suffer in quality. Personally I think the former isn’t much of an issue, but I’m not entirely unconvinced of the latter.
Google has been consistently clear that it doesn’t penalize AI-generated content so long as it provides unique value to searchers. If it’s not AI slop and it’s genuinely helpful you’re good. If you’ve used AI to write compelling content that layers in real human expertise and experience alongside the model-generated text that’s tailored for your audience you shouldn’t have much to fear.
I do see an argument for the quality complaints, though. As stated in their support document mentioned earlier, Claude’s watermarking works by choosing specific words in no particular order that if analyzed as a whole create a pattern that’s identifiable to machines yet unrecognizable to humans. But if wording is being chosen to fit a pattern, wouldn’t that skew outputs to more homogenous language choices?
Again, I think this is more of a problem for low effort content and less of an issue for people who take the time to train their AIs how to write effectively for their audience. Those who enter a basic prompt and run with whatever the output is were already at a disadvantage, and this change could add to that disadvantage. As for those of us who are trying to leverage AI to create highly valuable content, maybe this means we need to be extra diligent in our review of writing outputs to ensure that we’re not contributing to the Sea of Sameness. Can’t be upset about that one bit.
A new report from AI tracking software Promptwatch noted in mid August an 86% drop in the number of ChatGPT responses citing Reddit that has been holding steady for a couple weeks.
Several technically-minded SEOs are proposing that this shift could have something to do with a recent change in the way in which ChatGPT searches for answers. Suganthan Mohanadasan has been doing some excellent research around the changes to ChatGPT’s search tool and discovered that ChatGPT is now including a shortlist of known entities to research along with its normal web search function.
In his example he asks ChatGPT for a recommendation on the best AI note taking app. By inspecting the source code of the conversation, he was able to see that the model made searches for specific brands like Granola, Notion and Otterly before fetching any web results.
Suganthan’s articles are seriously worth a read if you want to learn more about how everything works. He’s quick to point out that as it relates to the recent Reddit citation drop his findings are correlational and acknowledges that there could be more to it than just the change to ChatGPT’s search tool.
However, this change could pose another substantial hurdle for less established brands. If ChatGPT (or other AI tools) are integrating brand shortlists based on their own training data into their responses, it might become much harder to break into the model’s recommendations over bigger players.
Zooming out a bit, this is another reason for businesses to focus on more holistic brand building and ignore some of the cheaper, quick-win tactics of the day that are only aimed at superfluous visibility. These AI models can change how they recommend brands quickly, and one strategy that’s always going to be a good investment is one that endears your business to your audience.
Updates by Paid Media Director Casey Anderson
In the past months, Google has been making big changes to put their Demand Gen campaign product into the forefront of its advertising platform.
They announced not too long ago that Display, as we have come to know it, will be deprecating later this year and that display ads will then run inside Demand Gen campaigns. The good news with this transition is that they’ve made it relatively easy to implement and there isn’t a requirement to use the YouTube placement. With Demand Gen being the only campaign type that allows “similar” audience segments, advertisers will again have this option in their display prospecting.
To make this even more appealing to advertisers who might not be familiar with Demand Gen, Google also lowered its budget recommendation for Demand Gen campaigns. Where it once was recommended to spend $100/day or 20X your target CPA to start, the new daily budget recommendation from Google has decreased to 10 times your tCPA. This change gives smaller advertisers who might have skipped out on Demand Gen in the past a chance to compete in these auctions.
On top of all of this, Google is offering a “Branded Search” conversion type that is available for YouTube and Demand Gen campaigns, making it easier to understand the search lift when brand queries have been made after viewing a video. This “consideration goal” will monitor interest over a 7-day period to help track engagement and effectiveness.
Google has instated a new 7-day offline conversion upload limit. Any conversions uploaded more than 7 days after the conversion occurs will still appear in standard reports, but will be ignored by data-driven attribution calculations.
This is an important update for advertisers who use offline conversions in their ads because upload timing can make a difference in measurement quality. It’s just another reminder that good data needs both accuracy and timely delivery.
Meta AI now can connect advertising campaigns and Google Workspace (Gmail, Calendar, Docs, etc.). For advertisers, this may assist campaign analysis and deeper reporting, including output in Google Slides and Sheets.
Meta AI uses the connection between assets to get more business context and provide better answers around paid social performance. This conversation-based AI tool will be able to identify creative opportunities, budget optimizations and reporting insights, moving beyond general assistance towards full workflow transformation.
Developing trustworthy architectures for insights with AI; Measurement Protocol replaced by Data Manager; and lots…
AI can speed up marketing data analysis, but it might lead you off a cliff.…
GSC now shows views for third-party platforms like IG, YouTube & TikTok; Yelp partners with…