AI outputs can be wildly inconsistent, and Rand Fishkin recently spotlighted this issue. His research revealed that AI tools produce varied brand recommendations, which highlights the need for a deeper understanding beyond ranking positions.
After reading his work, I realized the solution is rooted in something I’ve been developing for years – building consistent visibility through confidence and corroboration.
Fishkin’s data showed that AI systems are confidence engines. They draw results based on confidence levels, which explains the inconsistency in output. It’s a problem when there’s low confidence, but once AI systems are confident, they provide consistent recommendations.
The journey to AI confidence involves several stages, and understanding this process can fundamentally change how brands approach AI visibility.
Take the entity home as an example. It’s the foundation of AI interpretation of your brand. Confidence also builds when third-party data aligns with your own narrative. Brands that manage this well don’t just appear in AI recommendations; they dominate them.
There’s a method behind all this that I’ve formalized and even filed for patenting. It’s a complex system of strategies but starts with ensuring that your brand’s digital footprint aligns perfectly with high-authority sources.
Fishkin’s work confirms the importance of AI visibility, a subject I’ve been tracking and developing solutions for over the last decade. It bridges a significant gap in understanding how brands can leverage AI for long-term authority and presence.
I’ve discovered that Google Ads now offers ready-to-run experiments directly within the Experiments page, making it easier for me to test optimizations quickly without a complicated setup.
These suggested experiments are based on my account’s setup and performance data, helping me uncover new ways to enhance results.
How it works: The platform provides suggestions for testing various bidding strategies, creative variations, and new campaign features, all accessible right in the Experiments dashboard.
Every recommendation comes with a pre-configured setup, so I can either launch them immediately or adjust the settings to better fit my needs. These suggestions are conveniently displayed alongside the standard Create Experiment option, streamlining the process.
Why I care: Google’s effort to simplify experiment setups significantly decreases the time and effort I need to put into testing. It allows me to act swiftly on optimization ideas and maintain a consistent flow of improvements. However, I still review each test configuration to ensure it aligns with my campaign goals and doesn’t lead to unnecessary resource expenditure.
Zoom in: For instance, I might see a prompt suggesting I enable final URL expansion to boost campaign performance. These recommendations appear as pop-ups inside the Experiments interface, guiding my decisions with relevant insights.
The big picture: Google is embedding more automated guidance into Ads workflows, nudging me towards continuous testing and pursuing data-driven optimizations.
First seen:This update was first spotted by PPC News Feed owner, Hana Kobzová, shedding light on these helpful enhancements.
I recently embarked on a fascinating exploration of ChatGPT’s brand recommendation patterns, and let me tell you, the findings offer a lot to chew on!
We all know that AI responses are a roll of the dice – ask the same question ten times, and you’re bound to get ten different answers. But I couldn’t help but wonder, just how varied are these responses?
Rand Fishkin’s intriguing research dives into this very question. His findings have significant repercussions for how we approach AI visibility tracking for brands.
Fishkin experimented with prompts ranging from recommendations for chef’s knives to cancer care hospitals, as well as Volvo dealerships in Los Angeles.
His results showed that AI systems like ChatGPT almost never recommend the same set of brands in the same order twice.
Moreover, when asking about something specific like running shoes, certain brands tend to appear more frequently than others.
Building on this research, I zeroed in on B2B scenarios, adding some of my own twists: does the complexity of the prompt or the competitiveness of the category make a difference to AI’s consistency?
To investigate, I crafted twelve varied prompts, half of which addressed highly competitive B2B software categories, like accounting, and the rest focused on niche categories, such as user entity behavior analytics (UEBA) software.
Further, I examined simple prompts against nuanced ones that included specific personas and use cases.
Each prompt was fed into ChatGPT 100 times using different IP addresses to mimic 1,200 unique users.
Now onto the juicy part: the findings.
Submitting a single prompt to ChatGPT 100 times revealed that, on average, 44 different brands got mentioned. However, some response sets listed as many as 95 brands, heavily dependent on the category.
Notably, competitive categories yield twice as many brand mentions per 100 responses compared to niche ones.
Simple vs. nuanced prompts? ChatGPT typically mentions fewer brands in response to nuanced requests, but this isn’t a hard and fast rule.
When diving deeper into ChatGPT’s brand consistency, I found that in a set of 100 B2B software recommendations, only about five brands (11% of the total) were mentioned 80% or more of the time.
Dominant brands in a category like accounting software were names we all recognize: QuickBooks, Xero, Wave, and the like.
If you’re not among the big guns, working within a niche offers a strategic advantage given the increased chance to be consistently recognized by AI.
For marketers, this study underscores the necessity of standing out and perhaps carving a niche if dominance in a broad category seems out of reach.
Moreover, most AI visibility tools might not give you the full picture if they’re conducting only a single spot-check. For more reliable data, multiple runs per prompt are essential.
So, if you’re tracking pivotal prompts, run each a handful of times to get a better sense of your brand’s visibility.
I’m excited to share that future reports will explore ChatGPT’s understanding of brands and whether consistent recommendations reflect deeper brand awareness.
This article was originally published on Visible and republished with permission.
