“All I know is that I know nothing.” This quote (along with variations of it) is frequently attributed to Socrates, via Plato’s account of him.
For most of the past year, it’s also what the leader of almost any company specializing in AI search would tell you.
If you asked one of these companies to explain why some content surfaces more frequently than other content and why different LLMs skew toward different sources, for a long time, they all gave you the same answer: No one really knows.
But that hasn’t stopped dozens of companies from building their businesses around the practice of helping brands show up more often, and sometimes more positively, in AI search.
The practice (which is interchangeably referred to as generative engine optimization, GEO, and answer engine optimization, AEO) has taken off in the past few years, despite the fact that the major LLMs “are black boxes,” Aniket Deosthali, CEO of AEO company Envive, told AdExchanger. As in, “no one knows what’s happening” within them.
Keeping things fresh
While no one can explain with 100% certainty why some sources and content consistently outperform others, some best practices have made themselves clear – and AEO experts are starting to shine a crack of light into these previously impenetrable boxes.
For one, it’s crucial for consumer-facing brands to show up on third-party sources, like listicles and affiliate articles, according to Gintare Rimolaityte, chief commercial officer at AI search company Trendos. The self-serve analytics platform, which monitors prompt results on an ongoing basis, has found that different industries seem to value different sorts of content, according to Rimolaityte.
Apparel brands, for instance, need to have “well-structured” content on their own website (like Q&A formats), because the brand’s own site is “what AI engines believe is authority,” she said. Consumer finance brands, on the other hand, need to rely almost entirely on how they show up in third-party sources.
Experts also recommend that brands produce fresh, up-to-date content on a regular basis, said James Cadwallader, CEO and co-founder of AEO company Profound, which uses a proprietary model to track LLM prompts to help its clients determine the types of queries they want to show up against. “The citations that were informing a category will sort of drift and be replaced by new citations, new sources,” he said.
Plus, older content isn’t necessarily formatted in a way that’s desirable for AI engines. AI citations skew toward “highly schematic content,” said Cadwallader, like comparisons between different brands, the heavily debated listicle, etc.
Indeed, “human-driven, human-written content that is optimized for AI is absolutely the way to go,” according to Leah Nurik, CEO and co-founder of Brandi AI, yet another AEO company, with a focus on brand intelligence.
Any brand aiming to “future-proof” itself and really connect with its customers shouldn’t apply “tricky tricks to trick the AI,” she said. (Try that one five times fast.)
Play stupid games, win stupid prizes
There are a lot of “tricky tricks” out there, which is to say, practices that are less likely to work, ones that might lose trust among consumers or ones that go against unspoken rules of the internet.
For example, the rampant Reddit-ification of AEO (i.e., the practice of increasing brand mentions on Reddit, since it’s a source ranked highly by LLMs) “violates the most sacred principles of Reddit,” as Shafqat Islam, president of AI marketing platform Optimizely, put it.
The degree of trust built by community members is what historically made the platform “weird and wonderful,” said Islam.
Although it’s beneficial for brands to increase their mentions on Reddit, it’s “a little bit sad” to see the platform grow less community-driven, Rimolaityte agreed.
Reddit is skeptical of the assumption that a heavy footprint on the platform equates to better performance in AI search, a Reddit spokesperson recently told AdExchanger. The company “[questions] the legitimacy of any provider suggesting” they can guarantee LLM performance for a brand through Reddit posts, the spokesperson added.
Recently, Reddit has begun to push back internally against this “stealth marketing content,” per Bloomberg, with its own spam-reporting AI tools.
But Reddit-maxxing, as one might call it, may be losing its AEO appeal anyway.
AI models are smart: Once brands find a loophole to show up more in search, LLMs often catch on quickly.
“Just look at the history of SEO,” Nurik said. As soon as Google saw that people were employing new tactics to show up in search, it began “to penalize the people that were trying to game the system.” Now, the same is happening in AEO.
Prompt Watch reports that Reddit’s share of citations in ChatGPT dropped from about 3.83% from July 18 through August 7 to around 0.5%, where it sat in mid-August.
The first big drop was on August 8 – the same day that ChatGPT changed its search behavior to focus more on pulling information from specific websites, rather than scanning the whole open web. The second was on August 14, for no discernible reason.
One prevailing theory behind this citation drop is that the AI models have caught on to human AEO strategies and are now viewing platforms like Reddit as less reliable sources.
The fight for what’s right
Other techniques to increase AI search visibility, however, are just starting to gain traction among advertisers.
For instance, there’s been a recent surge in AEO vendors contacting publishers – AdExchanger included – directly and asking them to edit existing content to mention a specific company, a tactic that, one could argue, challenges the integrity of traditional journalism.
But the companies that employ this tactic don’t see it that way.
Since most of the content they’re requesting changes to is either evergreen or affiliate, it isn’t “the type of journalism” that “you would want to maintain the integrity of,” said Rahul Jain, CEO and co-founder of Noble, a company whose entire business model is reaching out to publishers on behalf of its clients to make such edits.
Noble’s process is fully automated but is designed to make the outreach feel a bit more personal. It emails publishers under made-up names (of various genders and ethnic backgrounds), requesting that their client be added to a specific article on the publisher’s site.
Noble determines which article would be a good fit for a given client by analyzing a list of LLM citations produced by Profound (with which Noble has an integration). The list is organized by prompt.
This inevitably means that Noble’s clients could end up in competition with one another. For instance, if Noble has several clients in the AEO space, its system might flag an article that shows up frequently when LLM users search for the best AEO platform and automatically email that publisher multiple times, each on behalf of a different client.
Noble has only been around for a year and is “still trying to figure out” the best way to approach publishers using this model, said Jain. It isn’t “this big established company with big pockets trying to change how journalism works,” he added.
Its response rate from publishers is about 4% to 10%. Not all of those responses are agreements to revise their existing content, of course; some of them are presumably “Hell no,” or, as AdExchanger wanted to know, “What is this?”
For some clients, the response rate – and the number of yeses – is worth the cost.
For others, not so much.
A C-suite exec at one ad tech platform, who spoke on the condition of anonymity, told AdExchanger that their company withdrew from Noble after only two weeks, having received “pushback” from the publishers that Noble was contacting on their behalf.
“I don’t want to tarnish my brand by doing spammy outreach,” they added.
Back to basics
And Noble is far from the only company doing this sort of work.
Direct outreach to publishers could be the next Reddit – or the tactic that everyone gloms onto until it loses its luster and the next shiny thing appears.
Still, what consumers perceive as “good” content – engaging, well-researched and catered to humans – hasn’t changed, said Optimizely’s Islam. “Just do all the SEO stuff, and you’ll be fine,” he added, as long as the structure of the content is easily digestible for bots – echoing Rimolaityte.
Trying to manipulate the system is “not a good use of your brain, and it’s not a good way to actually retain customers” who are, writ large, already losing trust in brands, said Brandi’s Nurik.
The best thing marketers can do, she added, is to “just do what your mother told you to do when you went to kindergarten: Be honest, live your truth, connect freely and authentically with people.”
