A few days ago, I posted a message in our team Slack that was equal parts venting and genuine frustration: “Claude Opus is bad and they should feel bad.” The immediate wave of replies confirmed my suspicion. A tool that previously worked like a seasoned analyst on demand had suddenly morphed into an overly confident sophomore, spinning up rogue sub-agents, racking up API fees, ignoring direct instructions, and completely losing track of the goal.
While we laughed about it, the exchange pointed straight to an uncomfortable reality every team needs to address: the tools, workflows, and models powering your strategy actively shift beneath your feet.
When a new system or model gives us great initial results, our natural instinct is to lock it down. We write the standard procedure, celebrate the time saved, and get comfortable. We assume that because the engine ran well during launch, it will keep running smoothly forever.
It won't.
Behind the scenes, AI models push unannounced updates and alter background parameters that degrade output quality overnight, while paid media algorithms shift and team structures designed for last quarter's volume choke under new demands.
Settling into a workflow just because it worked early on kills progress. Feeling comfortable is misleading. The moment you stop testing your setup is the exact moment your competitive advantage begins to fade.
Continuous improvement requires a mindset of constant evaluation. It means regularly reviewing your software, testing new models against default options, and replacing tools that no longer deliver top results. Past performance is only historical context.
Stay curious, test constantly, and keep your process sharp.
Patrick Patterson
CEO at Level Agency
What's interesting this week
Level lands two shortlist spots at the 2026 US Search Awards Level was shortlisted for Best PPC Campaign, for a value-based bidding overhaul built with Ancora Education, and Best Use of Search in Finance, for a three-site SEO consolidation built with IRA Financial. Both campaigns are detailed in Level's announcement post.
Key takeaway: Two shortlist spots in one year, across two different disciplines, says something about the breadth of the team and the clients behind them. Congratulations to everyone involved.
LinkedIn's "seems like AI slop" button gets clicked a million times Chief product officer Hari Srinivasan says more than a million members have flagged posts as AI slop since the button launched on July 30. Content flagged this way is now getting 40 percent fewer views. The button followed a Pangram Labs study that found 41 percent of LinkedIn's longform posts were fully AI-generated, the highest rate of any platform measured.
Key takeaway: Anything published under a brand's name on LinkedIn is now competing against a reader-trained detector, not just an algorithm. Run your own longform posts through the same instinct before they go out.
The IAB is building a framework to measure ads served to AI agents A framework due out November 12 will tackle how to credit ads when an AI agent, not a human, reads the page and makes the purchase decision. IAB vp of AI Caroline Giegerich says the working group's hardest task so far has been agreeing on anything at all. That starts with what counts as evidence an ad influenced an agent's decision.
Key takeaway: There's no standard yet, so don't wait for one. Start asking your platforms now what evidence they can already show you about agent-driven exposure.
What 60,000 display and video campaigns say about a channel everyone underrates
WordStream's first display and video ads benchmark report drew on data from over 60,000 LocaliQ campaigns, and the numbers explain why this channel gets dismissed by teams chasing search-style conversion metrics.
Education and Instruction posted the highest click-through rate of any industry at 0.40 percent. Beauty and Personal Care carried the highest cost per click at $5.58, reflecting how competitive that vertical has become.
The more useful numbers sit outside the industry table. High-impact display ads on the USA TODAY Network average 62 percent viewability. Video ads average 71 percent viewability with a 67 percent completion rate, even though their click-through rate sits at a modest 0.09 percent.
Microsoft's Navah Hopkins makes the point worth sitting with: display shouldn't be held to search benchmarks because it isn't trying to do search's job. Clutch's Greg Carter adds that most people don't click a display ad. They come back to the site directly, weeks or months later, when they're ready.
Key takeaway: For teams that measure display success by click-through rate alone, this report is a case for redefining the scorecard before touching the creative.
Google just admitted AI search is costing publishers traffic, and built a workaround
Google rolled out a "Preferred Sources" button that publishers can now embed directly on their own websites. It lets readers flag a site as one they want surfaced more often across Search, Discover, and Google News.
It's an expansion of a feature Google introduced in AI Mode and AI Overviews back in May. The timing matters: this is Google acknowledging, without quite saying so, that AI-driven search is sending fewer clicks to the open web.
The number that stands out: over 345,000 unique sources had already been selected as preferred through the AI experience version of this feature as of May. Google says people are twice as likely to click through to a source they've marked as preferred.
Alongside the button, Google is rolling out natural language feed customization in Discover. Readers can tell the app what they want more or less of instead of relying purely on algorithmic guesswork.
Key takeaway: Earning a spot on a reader's preferred source list is starting to function like a loyalty program for attention. It rewards the publishers readers already trust before AI ever entered the picture.
Adobe Podcast uses AI-powered audio processing to turn recorded voice tracks into broadcast-quality sound. Features like Enhance Speech remove background noise and room echo, while Mic Check analyzes recording environments to optimize input quality before production begins.
For teams producing podcasts, webinars, or video narration without full studio infrastructure, it significantly reduces editing overhead while maintaining professional standards.
Your AI visibility, scored.
When did you last audit what AI thinks of your brand?
Don't rely on an unreviewed algorithm to manage your brand’s reputation. Our scorecard pulls back the curtain on how AI platforms frame your brand sentiment and which sources they use to vet your authority.
Inside your custom report:
Visibility score and share of voice: The percentage of AI prompts your brand appears in, benchmarked against your top competitors across all major models.
Sentiment and positioning: Whether AI describes your brand favorably, neutrally, or negatively, and whether you're showing up as the primary recommendation or an afterthought.
Optimization roadmap: The specific content structure and authority signals you need to improve your score.
