As LLMs give increasingly personalized responses, it’s getting harder for companies to rank in AI search results. That was the central message of Devesh Khanal’s session at Content Marketing World last week.
Khanal, co-founder of SEO and video content agency Grow and Convert, has spent years advising clients through changes in Google and AI search. He says personalization has been a challenging shift to manage.
“A lot of the conversation about AI search is about citation weight and how Google lists fan out queries,” he explained. “We think this personalization hits content strategy for brands more than anything else.”
He saw this play out with one client firsthand.
AI Personalization in Practice
The client, Nancy, is the CEO of a business financing company. Grow and Convert asked how she first heard about them and why she wanted to hire them. The answer: ChatGPT.
The team also asked which prompts she typed into the chatbot. Here’s the prompt she used: “What are the best SEO companies in your opinion who have driven massive results and endless leads?”
She didn’t specify what her business is or the industry she works in. ChatGPT already knew.
“When it recommended us, it continued to bring specific stuff about Nancy’s Google business into its recommendation,” Khanal said. “High-intent content working for lead gen and real case studies. On top of that, the last detail you should pay attention to is that it recommends SEO strategy.”
Grow and Convert has been around for 10 years, and it built its brand around SEO. Since ChatGPT concluded that Nancy’s business needs SEO, its recommendation makes sense.
But Khanal wanted to test it further.
Where SEO and GEO Differ
After opening a logged-out, incognito ChatGPT tab, Khanal typed in the exact same prompt Nancy had. The response he got was completely different from Nancy’s. To help the LLM out a bit, he added terms like “business loans” and “funding space.” Still, ChatGPT did not recommend Grow and Convert.
“What’s going on? As marketers, what do we do?” he asked. “I have a prompt, I typed it in, and I can’t track it. This story elucidates this concept that we call personalized GEO.”
GEO, or generative engine optimization, is the practice of earning recommendations from AI tools. Because those recommendations depend on who’s asking, ranking for keywords on Google is no longer enough. Khanal named two key differences between SEO and personalized GEO to note.
- Prompts are conversations. “They’re not just single prompts, they’re conversations,” he said. “The context the LLM has from previous chats with the user dictates the answer, not just the prompt.”
- Brands need to be the recommendation, not background reading. For 20 years, content marketers aimed to get their brand listed among the many results Google used as background reading, and that worked. An LLM often gives a single recommendation instead, so the goal now is to be it. That changes what content needs to do.
Content Strategy as the Solution
Over time, content marketing has become accustomed to keyword-led research. As a result, SEO thinking has been continuously shoved into AI search.
“That’s why we’re talking about brand tracking and citations,” Khanal said. “We’re used to showing up as a blue link on Google.”
That’s where content strategy steps in. Content marketers can’t predict the one-on-one conversations that LLMs have with users. Instead, they have to know their clients’ business from the inside out.
“If we can’t predict beforehand what people are typing in and you can’t see that data in AI search, at least we can say we know our customer,” he said. “We need to make sure the content on our site covers these boxes and use cases.”
Content marketers can cover their customers’ needs and use cases by doing the following four things.
- SEO blog content. Write about specific features and use cases, targeting normal SEO keywords. Google traffic has declined, but it isn’t zero, and the same content can teach LLMs what your product does well.
- Case studies. LLMs cite them because they contain personal, specific details that personalized responses draw on. Khanal advised ditching short “problem, solution, big percentage” write-ups in favor of detailed ones about customers and their circumstances.
- Your marketing site. Skip vague, flashy feature pages. Explain features, use cases, and scenarios in detail, even if that means adding secondary pages.
- Offsite. Check your visibility data to see which sites LLMs actually cite for your topics, then focus your outreach on those sites.
Don’t Skip Your Own Site
Khanal was blunt about priorities. Three of his four tactics happen on your own website, and he disagrees with those who say onsite matters less. A few Reddit comments won’t compensate for thin content on your own site.
For offsite work, he said to go where your audience is. One of his clients had little presence on Reddit, because nobody discussed its industry on that platform. LLMs instead cited other niche sites, so that’s where the team focused its outreach. Across thousands of citations, Grow and Convert found that 86% of the citations for buying-style prompts came from sites that weren’t general-interest ones.
The main takeaway from Khanal’s session: be detailed on your own site, and save outreach for the specific sites LLMs already trust.
