It used to be enough for brands to just track what their human customers were saying about them.
Now they need to know what LLMs are saying about them, too.
At the same time, the rise of AI has sped up every factor of digital marketing. Campaigns can be designed and launched in a matter of minutes, and instead of looking at weeks- or months-old data, marketers want to see real-time insights.
Spotlight, a startup that helps brands understand and improve the ways they’re showing up in AI search, recently partnered with Cheil UK to help the agency give its clients more and faster insights that they can use to stand out in a new era of search.
By the AI, for the AI
Spotlight’s entire platform was actually built by AI agents – and that helps when having to make speedy product updates.
“We don’t want any humans touching the code,” Spotlight CEO Michael Hermon told AdExchanger.
If, for example, the Spotlight team were to wake up tomorrow morning and find that Google had changed something about AI mode in the UK, Hermon said, then “we have to immediately change the product to match that.”
In other words, there’s no time for meetings and a bunch of back-and-forth emails, because the updates need to roll out within hours. AI can do it in 20 minutes, he said.
Spotlight also takes suggestions and makes those changes rapidly, said Cheil UK CEO Chris Camacho.
One of Cheil’s brand clients, for example, observed that it would be helpful to have a way to track how prompt results evolve over the course of weeks or months, Camacho said, rather than only getting a single “snapshot in time.”
So Spotlight created that feature. Brands can now generate reports that show how they rank in response to a specific query over time.
After noticing how often chatbots tend to cite Reddit, for instance, one of Cheil’s clients started adding more comments about its brand to the site and was able to track the resulting increase in search mentions.
Thank you, next
And speaking of time, Spotlight takes the phrase “real time” almost alarmingly literally.
It scans the web hourly, according to Hermon, using APIs for models, including ChatGPT and Gemini, to seek out new tips and hacks posted on social media that marketers have said helped them increase visibility in LLM searches, like revisions to their HTML or Node.js script.
Spotlight takes these suggestions with a grain of salt. After all, not everything you read on Reddit, for example, is 100% true. It has one of its AI agents run an A/B test on a wide range of content to see if any given new technique actually improves visibility in LLM results.
If it does, Spotlight rolls it out to all users. If not, the agent trashes that idea and moves on to testing the next one – and again and again down the line, Hermon said.
To each their own
Meanwhile, no two LLMs organize information in exactly the same way, said Hermon. Claude and Gemini, for instance, both favor content that is “grounded in science” and includes statistics and citations from scientific websites, whereas ChatGPT is apparently a big fan of Reddit.
Every model, therefore, “has its own flavor” when giving recommendations, he added.For marketers who remember ye olde days of SEO, this probably sounds familiar.
The major search engines, such as Google and Bing, “have a lot of commonalities,” said Camacho, but brands still need to make “slight alterations” to their content to cater to their particularities, such as including certain keywords or adding content on a certain platform.
But now, instead of keywords, brands are optimizing for prompts.
There isn’t any public database that tracks how often certain prompts are used, so Spotlight analyzes a bunch of different data sources to try and figure that out, including the Google ecosystem (such as its Keyword Planner and Search Console tools), LLM outputs and Chrome extensions.
Even taken together, these sources aren’t large enough to guarantee a complete understanding of what people are searching for, said Hermon, but it’s sizable enough to give a “really good estimation” of how many people are using a given prompt.
And as more and more people adopt LLMs in their daily lives, said Camacho, it’s crucial for brands to understand how they’re showing up in LLM queries so they can “future-proof” their marketing strategies.
“Inaction,” Camacho said, “is actually far more risky than action.”