In the past, users looking for information might type in a question or a selection of key words to a search engine like Google or Bing. But now, AI search tools like ChatGPT, Claude or Perplexity are increasingly becoming the go-to source to find answers online.
As a result, marketers are racing to ensure their brand’s names show up in AI search — but there’s not yet much consensus on what the winning formula is. AI ‘visibility experts’ have sprung up to cater to this emerging market, as well as startups helping companies track their brand performance across various chatbots.
So how can you get started with generative engine optimisation? Sifted chatted to several startups to hear what’s currently working for them.
Advertisement
Get the basics right
- Make sure you have the technical SEO foundation right. Your site should be fully crawlable, without any “orphan pages” — those not linked internally to any of your other pages, says Adam Sidorczuk, a freelance SEO experts.
- Also remove unintentional no-index tags (meta tags which can block Google from indexing a page so it won’t appear in search results) or robots.txt directives (commands in a text file that tell search engine bots which URLs they can access on your site.)
- Implement Schema Markup on your website— aka code to help search engines understand your content better. Use structured data like FAQPage, HowTo, Product and Organisation to explicitly tell LLMs about the entities, attributes and relationships on your page, says Sidorczuk.
Understand where you’re showing up
- Look at where you are being mentioned. Which LLMs surface your brand? In response to which prompts? Observe the themes that come up and base articles or social media content around them, says Luke Costley-White, chief growth officer at DojoAI, a platform deploying agents to help companies with marketing tasks.
- Then, track performance. Measure whether actions like creating a Reddit thread or writing an article on those themes translate into traffic or visibility.
Tailor content to how LLMs and users behave
- Learn how the different AI bots source content. For example, Google’s AI overviews — those longer summaries you get from searching Google now — pull from video content a lot, says Lara Kennedy, the senior marketing associate at online recruiter Ivee (no surprise since Google owns Youtube). “Take just two minutes to record yourself speaking about a specific topic and upload it to YouTube.” ChatGPT, meanwhile, draws a lot from Reddit.
- Get into the minds of the chatbot users and think of the questions they’ll ask. Kennedy has rejigged Ivee’s website to try and ape some of the “conversational habits” of AI bot users. Ivee’s blog entries, for example, have been given chatbot-friendly headlines like “What companies truly support career returners?”.
Use the LLMs to help you
- Ask the LLMs. “How do you see my service? What sources do you find about it? How do you recommend services in my category?” It might sound strange, but LLMs often “reveal” their decision-making logic, and those insights can help you build a strategy, says Pavel Shynkarenko, founder of Mellow, a global payment platform for freelancers.
- Use AI to create content for you, but do it strategically. Instead of asking AI for full articles with none of your own input, use it to enrich your already existing content with UX improving elements such as Key Takeaways and FAQs. The great side effect is that you make your content easier to read for LLM bots, and you can include more keywords you care about on the page, says Sidorczuk.
Monitor — and iterate over time
- Track traffic with free tools. If your startup can’t afford AI visibility experts like Peec or PromptWatch to advise you on how to become “findable” on chatbots, no problem. “You can use Google Analytics to track traffic from different LLMs. You don’t really need another software to add to the stack,” says Kennedy.
- Don’t focus too much on one chatbot. OpenAI could eventually change the algorithm which will influence your performance on ChatGPT. Better to spread the risk and optimise for various models, says Marc-Pierre Hoeft, VP of client relations at Piabo.
