Tags
We know what AI can do; generate content, analyse data, recommend products, and take complex tasks off our hands.
The more interesting questions now sit around what happens when we let them: What should AI be trusted to do? When does its involvement need to be visible? And who’s responsible for judging the output?
Those questions are starting to show up in product decisions, platform policies, and regulation. That makes this month's developments interesting for reasons beyond another round of new AI features.
1. Google wants to do the first pass on your data
Google has introduced new AI features across Ads and Analytics, including new performance summaries, natural-language reporting, and competitor benchmarking.
The updates are designed to help users decide what their deserves attention, with Analytics able to reveal important changes, while Ask Advisor can help investigate them and Google Ads can turn questions into reports and explanations.
The expectation is that these changes will remove time spent hunting through dashboards, but it also changes how and what decisions are made by the user. If Google decides what looks unusual and suggests why it happened, teams need to know when they can accept the analysis, and where they need to challenge the interpretation.
2. LinkedIn draws a line at AI slop
‘AI slop’ has made it into LinkedIn's reporting options.
Users can now flag posts as ‘Seems like AI slop’, enabling LinkedIn to identify low-quality AI content. LinkedIn isn't saying that AI-assisted content is the problem, but it is responding to what happens when cheap production and poor-quality control impact what its users are looking for and expect.
3. ChatGPT Ads is looking increasingly like a serious performance channel
Only recently did ChatGPT announce ads were coming; now they’re adding conversion optimisation, attribution, product feeds, and testing multi-product carousels.
Unlike an ad interrupting a feed, the user may already have explained what they want and be actively narrowing down a decision. That's obviously appealing for advertisers, but it puts paid placement unusually close to advice people may perceive as impartial. Definitely one to keep an eye on to see how it develops.
4. Europe's AI transparency rules become real
The transparency requirements under the EU AI Act came into force on 2 August, bringing requirements around identifying AI-generated content and informing people when they're interacting with AI.
Transparency is shifting from policy into the work itself, because when content can pass through people, platforms, and models before reaching a customer, with AI doing anything from tweaking a sentence to producing the whole thing, that makes a seemingly simple question – was this made with AI? – more difficult to answer.
5. Anthropic tries to show where Claude was involved
Claude models launched in the EU from 2 August now support machine-readable marking, with watermarks in generated text and source metadata in files.
‘AI-generated’ is becoming a blanket description for a much more complicated creative process, so as transparency requirements grow, businesses need a better understanding of how AI contributed to something and how that needs to be shared with its audiences.
6. Meta joins the AI assistant party
Meta AI can now take an objective and work through multiple steps, including research, presentations, and recurring tasks, with connections to email and calendars for additional context.
Is this Meta’s attempt to catch up? Major assistants have already been moving towards more agentic ways of working so it will be interesting to see which platforms gain traction with users in the short term.
7. Google is letting your avatar do the talking
Google Vids can now use Gemini Omni to generate and conversationally edit video, while its new personal avatars can create a digital version of someone from a selfie and voice recording to deliver a script.
Repeatable videos, updates, and variations could be produced without another filming session every time something changes.
But with this, generated clips will be marked with SynthID. So the same product making video easier to create is having to make that creation easier to identify.
In the end...
We already know what AI can do. What’s becoming more interesting is what follows once AI is being used at scale. More convincing generated content opens up huge creative possibilities, but also makes knowing where something came from more important. Businesses have to think more about how much trust they place in the output and the conversation they are having with their audiences when it comes to transparency and authenticity.
