This is the second edition in trying to help you figure out how to stand out in LLMs. Over the last few weeks, we’ve spent time with dozens of startups navigating these shifts. Some already adapting, some still guessing. This issue shares what we’ve learned so far and what we’re still trying to figure out.

TLDR of the newsletter:

1. You should care about Bing more than ever
2. GEO means prompts matter more than keywords
3. Most citations come from off-site content, not your website
4. LLMs reuse your copy, so write clearly and quote-ready
5. Small brands can win by studying what LLMs already quote
6. User-generated content is highly quotable when it’s fresh and detailed
7. Expect less traffic but more purchase-ready visitors
8. GEO is still evolving and there’s a lot we don’t know yet

We’re moving from keywords to prompts, from clicks to answers, from traffic to citations. Generative AI tools like ChatGPT, Perplexity, Gemini, and Claude are changing how people discover and decide. Bing is suddenly relevant again.

This shift means:

• GEO (Generative Engine Optimization) is the new SEO.
• Your content gets summarized, not just indexed.
• Fresh UGC and off-site citations matter more than ever.

This newsletter is long, but I promised you that we would get into the weeds. Dive into the sections that interest you, and if you have questions or feedback, reply and let me know.

Why You Should Care About Bing

Search behavior is changing. People used to Google short phrases (an average of 4 words), but now they’re asking full questions (23 words, on average) and expecting direct answers. That shift is massive. It moves the point of discovery from your homepage to a few quoted lines inside ChatGPT or Google AI Overview.

And here’s the punchline: most of those lines don’t come from your site.

They come from Reddit threads, review snippets, niche blogs, and yes, Bing-indexed pages.

So even if Bing hasn’t mattered to you for the last decade, it matters now. Because Perplexity and ChatGPT rely heavily on Bing. And they’re leading this new wave of zero-click commerce.

What Is GEO? Ask the Models Themselves

We asked leading LLMs to define GEO in one sentence:

• Perplexity: GEO is the strategic process of optimizing content so it is more likely to be referenced or cited by AI search engines.
• ChatGPT: GEO is a new branch of SEO that enriches content so AI engines preferentially cite your pages, amplifying brand visibility.
• Grok: GEO is optimizing content for visibility in AI-driven search results.

They agree on the goal: help LLMs find and quote you. Where they differ:

• ChatGPT pulls from Bing and training data. Get mentioned in trusted sources, and you’ll show up.
• Gemini relies on Google Search. Classic SEO + E-E-A-T still wins.
• Claude is static. If you weren’t visible in 2023, you’re not in the mix.
• Perplexity uses real-time Bing. Fresh, structured content gets cited.

What’s Changed in GEO (vs. SEO)

1. Prompts > Keywords People ask real questions now. GEO rewards answers, not just keywords.

✅ Write with depth and clarity. Think like a support agent, not a keyword machine.

1. Mentions > Clicks Your product might be cited with no visit to your site.

✅ Track brand mentions inside AI answers, not just traffic.

1. AI Reuses Your Voice LLMs summarize your copy. Make it quotable.

✅ Use clear, specific phrasing. Write like you want to be quoted.

1. No Transparency You might show up, but won’t always know why.

✅ Start testing and tracking GEO visibility. Manual or with new tools.

What’s Still True

• Structured data still matters
• Quality content still wins
• Authority still helps
• Good UX still matters
• Reviews still sell

What’s Probably Not Going to Last

• Ghost Content: Hidden pages for bots. Risky.
• Prompt-Stuffing: Awkward, clunky. Models don’t need it.
• Cloaking: Show bots different content. Big no.
• Gaming LLM Training: Flooding the web to bias models. Doesn’t scale.

Smarter path: Publish helpful content where users and bots both benefit.

Field Notes: 20 Hours With GEO Startups

• The average prompt is 23 words, not 4. Context is everything.
• Winning content is 100-150 words, structured, Q&A-ready.
• Top-of-funnel, mid-funnel, and bottom-funnel prompts are all different. You need to cover them all.
• ChatGPT o3 is the best for prompt discovery. Ask what your audience is likely to ask.

Citations Matter More Than Your Site

70 to 85% of quoted passages in LLMs come from off-site sources. Your site contributes only about 25% to GEO visibility.

✅ Build presence in:

• Reddit, niche forums
• Third-party reviews
• Influencer blogs

New GEO KPIs to Track

• Visibility Score
• Prompt-Win Rate
• Citation-Share-by-Engine
• Product-Tile Frequency
• Sentiment Delta

✅ You’ll probably need external tools. It’s a numbers game.

Traffic Will Drop. But Buyers Get Hotter.

GEO drives zero-click commerce. Less traffic, higher intent.

• CTR drops ~34.5% with AI Overviews
• Visitors that do click are more informed, more likely to convert

Bottom line: you will see fewer visits, but better buyers.

Small Brands: Don’t Panic. Just Listen.

Most AI answers feature big brands. But you can still win.

✅ Study what AI quotes. That’s your cheat sheet.
✅ Publish content in the formats and forums LLMs trust.

UGC Is Quote Bait. Use It.

LLMs love user reviews. Especially recent ones. Especially detailed ones.

✅ Refresh reviews often
✅ Highlight specifics
✅ Mark up with schema
✅ Encourage off-site reviews
✅ Turn UGC into structured Q&A blocks

Quote-worthy: “These sweatpants are perfect for chilly mornings. Super soft and warm without feeling heavy.”

Still Unknown:

There are still big gaps in how GEO actually works. Here’s what we’re watching:

• How answer agents rank products: LLMs are starting to act more like agents: surfacing product cards, prices, reviews, and links. But how they pick which product to show isn’t clear. Is it freshness? Reviews? Site speed? No one knows.
• What authority means in GEO: In SEO, we had Domain Authority or PageRank. In GEO, there’s no clear proxy. Some say citation frequency, others say brand familiarity. But there’s no consensus, and no metric to measure it.
• Why identical prompts give different answers: Ask the same question across ChatGPT, Perplexity, Gemini, and Claude and you’ll get different brands, quotes, and formats. It’s likely due to how each model retrieves and ranks sources, but we still don’t know the weight of each factor.
• How personalization affects visibility: ChatGPT lets users set preferences. Perplexity uses search history. Gemini pulls from your Google profile. Personalization is clearly influencing answers, but we can’t see how, or measure its impact.

TLDR: There’s still a black box. But the outlines are becoming clearer.

We’ll keep watching, testing, and learning together.

Talk soon, 

Tomer