Google Overview fiasco, PayPal dataset monetized, Adobe trains AI with user content, causes uproar, YouTube at war with ad blockers and other??topics
Whatagraph
The fast and easy-to-use platform for monitoring and reporting on marketing performance.
Welcome to the July issue of Summarized, a marketing newsletter by Whatagraph, where we share our latest articles, case studies, top news from the digital marketing world, and platform updates each month.
Google address the failures of AI Overview
Through the end of May and entire July we have witnessed the spectacular catastrophe that was the deployment of Search's AI Overview functionality. Openly mocked across the internet, simple Search queries returned gems such as advice to use glue to make pizza sauce more sticky, to eat rocks "because they are good for you", the description of a "scissor kick" as if including runinng with scissors, or completely made up content such as a story about John Backflip (sic!) performing the first ever backflip in 1316 in Europe, before being exiled by his rival, William Front Flip, on accusation of witchcraft. Yes, that was a real search result for a "first person to backflip" Search query.
One of the causes for such nonsensical results is the recent decision from Google to train their AI on Reddit content. This was done under the concept of Reddit being the best source of "conversational", knowledge-seeking threads with true replies from highly experienced professionals. What Google failed to account was the amount of useless, facetious or outright intentionally misleading replies from online trolls Reddit is also well known for. This stark example shows the dangers of how disconnected large corporations and their staff are from the basic user knowledge and experience. Few reasonable users would trust the most upvoted Reddit answer to be the correct one, especially so far as to be presented as the sole and recommended answer, highlighted at the top of the first Search result page.
The following break of the month reaction from Google however failed to take accountability for the results provided to users. Instead, Google blamed "the information gap" or the AI "misinterpreting queries, misinterpreting a nuance of language on the web, or not having a lot of great information available", as if the fact that few people asking "How many rocks should I eat?" before is a good reason for the AI to suggest that we should eat a few daily as it is a healthy habit.
Google then claimed to have introduced technical improvements into their AI Overview system, such as better detection of nonsensical answers, exclusion of satire and humor content, limiting usage of user-generated content - which puts to question their recent acquisition of rights to use Reddit posts? - and stronger general guardrails around such topics as news or health. While those adjustments seem to be correct, if a tad late, they failed to stop AI Overviews from amusing users with further nonsense throughout June and the start of July.
How do you think Google can recover from this PR disaster?
PayPal is creating a new ad network based on its entire purchase dataset
In a quite surprising move, PayPal have decided to begin fully monetizing the dataset of hundreds of millions of users. To enact growth, the company has started an ad network fueled by data gathered from both the PayPal and Venmo apps, as well as PayPal enabled desktop browser transactions and the PayPal debit cards.
Those efforts will be driven by the newly hired VP and chief of the freshly established PayPal Ads division, Mark Grether. When interviewed by Wall Street Journal, he was quoted to say "If you’re someone who’s buying products on the web, we know who is buying the products where, and we can leverage the data. Consumers who use the PayPal credit card would also provide the company with data from real-world bricks-and-mortar stores."
All shoppers will have their data included by default, on an opt-out basis. PayPal aims to sell ads not only to its own customers, but to so-called "non-endemic advertisers", or those that don’t sell products or services through PayPal. Those companies might use PayPal data to target consumers with ads that could be displayed elsewhere - for instance on other websites or connected TV sets. The dataset to be used is surely extensive, as only in the first quarter of 2024 PayPal itself, without accounting for Venmo, has processed 6.5 billion payments by approximately 400 million customers, according to their latest earnings report.
Is this a new ad market worth expanding into or do you think the privacy concerns make the potential PR issues too risky?
Adobe's controversial move to use content to train AI
In a recent update, Adobe have altered the Terms of Use for key software such as Photoshop, Illustrator and other Adobe Creative Cloud apps. In these updates Adobe have granted themselves "the right to access user content through both automated and manual methods", using "techniques such as machine learning in order to improve our Services and Software."
To nobody's surprise this announcement has caused a tumultuous uproar from creatives, as well as much concern among their employers. Given Adobe's recent push towards generative AI in order to rival Midjourney and DALL-E, as well as OpenAI's Sora video generator, many suspect the updated terms will allow the company to leverage high-quality, user-created content to train future AI models. As most users utilize Adobe software for professional work, concerns instantly emerged, from questions about intellectual property, to outright fear of infringing upon NDAs and other confidentiality agreements by inadvertently "releasing" content being worked on into the AI wilderness.
The outrage is further motivated by Adobe granting itself the “non-exclusive, worldwide, royalty-free sublicensable, license, to use, reproduce, publicly display, distribute, modify, create derivative works based on, publicly perform, and translate the Content.” This implies an incredibly vague and broad range of rights which usually would be purchased by licensing the content - and would often be outright disallowed by aforementioned NDAs. The situation is not helped by the fact that this clause is opt-out for users and the change to the ToUs was quietly snuck in by Adobe back in February 2024.
Despite Adobe's reassurances and vague promises, users' creative work, assets and NDAs may have already been compromised, which may serve as a serious breach of trust - as there is no way to know now if and when Adobe breaks those promises.
How would your brand regain customer trust and good will after this kind of outcry?
YouTube at war against ad blockers
For over a year now, YouTube has been combating ad-blocker software. Initially, there were pop-up messages informing users that “ad blockers are not allowed on YouTube.” Then a three-strikes policy, warning viewers that playback would be halted after three videos, unless they disable their browser extensions. Some warnings even featured a countdown clock which could last up to 60 seconds, displaying to users how much time they had left to either "allow YouTube ads" or "try YouTube Premium,” before they are shown the message again.
Throughout this conflict ad-blockers have persisted and adapted, resulting in a perpetual tug of war. As the means to utterly break the creative attempts at either blocking the display of ads or skipping the so-called "sponsor blocks", YouTube have turned to directly server-side injecting ads into the video stream, re-rendering it on the fly. This method makes it a lot harder for software to discern where the actual content stops and an ad begins.
Currently, there is no foolproof remedy to this issue, but this is absolutely not to say it won’t emerge in the future. Filter developers are already working on short-term fixes and are actively looking for a more stable solution.
Do you think Google have finally won the YouTube ad-blocker war? Or is this just quiet before new developments emerge and ads on YouTube will once again be avoided by most tech-savvy non-Premium users?
What's new at Whatagraph?
In July we introduced updates to our widgets and report tabs features. Make sure to follow our Canny page to stay up-to-date with all updates!
Let's start with a couple of less noticeable but oh-so-nice updates to the widgets you use all the time:
领英推荐
Scrollable text widgets
No matter how extensive and detailed your text summary is, it will now occupy as little space as you wish!
This update is pretty straightforward: if you type more text than a widget can fit in its current size, a scroll bar will appear on the side.
Trend lines
You can now highlight a trend in the area, line, and bar widgets. To do that, add a line showing the general progression for a certain metric over time.
The trend line is off by default, but this new option can be found in a metric's settings. It's the most granular level of settings out there. There are also multiple trend line types to choose from.
Pie charts and donut charts update
By default, you see both numeric values and percentages when creating pie and donut charts.
Now, if you want to stick only to one (numbers or %), you can select a suitable option in the widget settings.
Take a look at this video walkthrough that explains the updates.
Hiding tabs when sharing Reports and Dashboards
With this update, you can mark a tab as "hidden," and it will not show when you share a report with clients or stakeholders.
To achieve it, open a tab's actions menu, and the new "Hide tab" option will be right there.
If you choose to "hide" a tab, it will no longer appear when:
(!) Please note that "unhiding" a tab will make it visible again if a report is opened using a live link you shared previously.
That's all for this month! If you enjoyed this issue, please follow us on LinkedIn and stay subscribed to the Summarized to receive more updates in the future.
The Whatagraph Team