Silicon Valley Girl Highlights The New AI Search Battle At
Marina Mogilko says the real battle for visibility has moved inside AI engines, pushing creators and brands to optimise for Google’s AI Overview and chatbots like ChatGPT instead of relying on traditional SEO alone Marina Mogilko/Instagram

Apple briefly removed Silicon Valley creator Marina Mogilko's podcast from AI‑driven discovery, she claims, as assistants like ChatGPT increasingly decide which businesses get customers. She says visitors arriving from ChatGPT convert almost ten times better than those from Google search and in a new Silicon Valley Girl episode shares a five‑step 'Generative Engine Optimisation' (GEO) blueprint she says helped her show become a regular recommendation inside ChatGPT and Google's AI Overview.

The host tells listeners that 'before they ever reach your website or blog, they're asking ChatGPT, Claude, Perplexity: "What should I buy? Who should I hire? Which company sells this?"' and warns 'the AI doesn't hand them 10 links anymore; it hands them 3 names, maybe 5. But if you're not one of them, the sale is over before it starts.'

She says internal data from her newsletter shows 35 per cent of people in the US now start shopping inside an AI assistant instead of a search engine.

AI Search and GEO Rewrite Discovery

Across the episode, Mogilko sets out a plan she calls Generative Engine Optimisation, or GEO, aimed at ranking inside AI chatbots rather than simply chasing classic search pages. She defines GEO as 'Generative Engine Optimisation, optimising whatever you are producing online, your content, your website, for those chatbots.'

To show what is at stake, Mogilko recounts a story involving New York real estate broker Ryan Serhant and a $50 million (£37.1 million) deal that almost collapsed after both buyer and seller asked ChatGPT about the price. Serhant has described the episode publicly, including at Fortune's Brainstorm Tech conference, where he said the deal 'nearly fell apart'.

The GEO Blueprint in Five Steps

Throughout the episode, Mogilko boils GEO down to a practical five‑step blueprint:

  • Ask AI to audit your site by pasting in your URL and questioning 'how visible are we in AI search, and why aren't we showing up?'.
  • Give crawlers static, readable pages by rebuilding key content as plain‑text HTML, including full transcripts and summaries on episode‑level URLs.
  • Fix your identity records in places AI reads most, such as WikiData and platform bios, ensuring labels like 'podcast host, entrepreneur, and angel investor' match across entries for your name and show.
  • Align bios and descriptions word for word on Apple Podcasts, Spotify and other platforms so every source repeats one identical, accurate sentence about your podcast.
  • Match titles to real AI questions and track them using tools such as Peak AI, which she says reveal 'the exact keywords a model searches for internally while it's answering questions' so episode titles can be rewritten around those phrases.

From Invisible Podcast To AI Recommendation

Despite interviewing leaders from LinkedIn, GitHub, Perplexity and Microsoft, Mogilko says Silicon Valley Girl initially failed to appear in AI recommendations for podcasts. 'After 50‑plus CEO interviews and the models acted like the show did not exist, isn't that wild?' she asks, saying ChatGPT, Gemini and Perplexity did not surface her programme when users asked about Silicon Valley podcasts.

That absence prompted her team to run 'a full GEO turn'. The first step, she explains, was to paste her site URL into Claude and ask 'how visible are we in AI search, and why aren't we showing up?', with the chatbot flagging missing HTML parameters and transcripts hidden behind JavaScript.

Mogilko's team rebuilt the podcast's web presence so each episode had its own static, crawlable page with a full transcript in plain text. She says exposing transcripts added '8,000 to 15,000 words per episode the AI can now read, and quote back with our name on it.'

Mogilko points to WikiData, which 'sits behind Google's knowledge panel and feeds Gemini', and notes that her own entry still described her as 'vlogger and a YouTuber', meaning 'a vlog never gets pulled into a podcast answer.' Her team changed the record to list 'podcast host, entrepreneur, and angel investor', added Silicon Valley Girl as its own entry, and translated the information into 11 languages.

Matching AI Questions and Tracking Wins

To ensure AI and search engines could find every episode, she submitted a sitemap through Google Search Console so pages were crawled and indexed.

To see whether the GEO work was paying off, she uses Peak AI, which 'checks the three engines we have turned on: Google, Google AI overview, ChatGPT, and Perplexity' and shows if Silicon Valley Girl appears in answers. She reports that the podcast's AI search visibility 'doubled', with some queries where 'we come up 34 per cent of the time.'

Writer's pick