I had a client last fall — a B2B SaaS doing payroll for restaurants — who was showing up in Google Gemini for their key terms but completely invisible in ChatGPT. That's the opposite of the usual problem, and it took me a hot minute to figure out why. The answer turned out to be embarrassingly simple, and it tells you almost everything you need to know about how Gemini works differently from the other AI models. I'll get to it later in this piece because it's a good story, but first let me lay the groundwork.
Gemini is not just another LLM chatbot. It's the AI layer of Google, which means it shows up in more places than ChatGPT does, it inherits Google's existing trust hierarchy, and it asks for a separate consent before it'll use your content for AI. If you treat it like "ChatGPT but Google," you'll miss the levers that actually matter. So let's walk through what I've learned in 18 months of clawing client sites into Gemini's answers.
The three places Gemini "is"
Before we talk tactics, you need to know that "showing up in Gemini" means at least three different things, and each one has its own dynamics.
1. The Gemini app at gemini.google.com. This is the standalone chat product. People go here on purpose to talk to an AI. It pulls from a mix of Gemini's training data and live Google search results, depending on the question. Citations link out. This is the lane most people mean when they say "show up in Gemini."
2. AI Overviews in Google Search. The big purple box at the top of Google search results that summarizes the answer before the blue links start. This is technically powered by Gemini under the hood, but the surface is Google Search. The user did not type into Gemini — they typed into Google. AI Overviews appear on something like 40% of informational queries now (the exact rate moves around) and they're a huge zero-click traffic siphon, but if you're cited in one, your domain shows up in the source list and people do click through.
3. Gemini inside Google Workspace. When somebody in Gmail or Google Docs uses Gemini to draft an email, summarize a doc, or research a topic, Gemini sometimes pulls from the web. The user is not searching Google — they're working in Docs. Your content can still end up in that summary if the model decides to pull it.
The work to win each of these is mostly the same, but the measurement and the click economics are different. AI Overviews drain traffic from your other Google rankings, the Gemini app drives qualified clicks, and Workspace integration is almost impossible to attribute directly. Plan accordingly.
The single most important lever: Google-Extended
OK, here is the thing most people get wrong. Google has two separate consent flags for crawling your site:
Googlebot— the regular search crawler. If you allow this, you appear in Google Search.Google-Extended— the AI training and AI-answer crawler. If you allow this, your content can be used in Gemini, AI Overviews, and Vertex AI products.
These are completely independent. You can be in Google Search and not in Gemini. In fact, that's the default state for any site that hasn't explicitly allowed Google-Extended. Google introduced this opt-in mechanism specifically so publishers could be in regular Search results without being used for AI training, which is a perfectly defensible product decision but it means most small business sites are accidentally opted out of Gemini.
You opt in by allowing it in robots.txt:
User-agent: Google-Extended
Allow: /
That's it. One block in your robots file, and you go from invisible in Gemini to potentially being cited in answers about your industry. I'd put this in the top three "free SEO wins in five minutes" alongside fixing a broken canonical and submitting your sitemap to Bing.
yourdomain.com/robots.txt. If you do not see Google-Extended mentioned, you almost certainly are not opted in. Adding it is harmless even if you don't care about AI — it does not affect regular Google Search at all.
Be in Google Search first, then everything else
Here's a thing that I cannot believe needs to be said but, based on my audit conversations, very much does: Gemini is built on top of Google Search. If you are not crawlable, indexed, and ranking somewhere on the first three pages for your target queries in regular Google, you are not going to magically show up in Gemini's answers.
This is different from ChatGPT, where you can be a tiny site nobody's heard of and still get cited because OpenAI's search index pulled you in based on content quality alone. Gemini inherits Google's authority hierarchy, which means it tends to cite sites that already do well in Google. The good news: the work for both is the same.
So before you obsess about Gemini-specific tactics:
- Make sure every important page is indexed in Google Search Console.
- Fix any "discovered but not indexed" or "crawled but not indexed" buckets.
- Get your Core Web Vitals into the green if they're not already.
- Ship structured data (more on which kinds in a minute).
- Make sure your sitemap is current and submitted.
Once you're solidly indexed and ranking, then the Gemini-specific levers start to compound. Before that, they don't matter.
The YouTube lever (the one nobody talks about)
Gemini pulls from YouTube transcripts heavily. Heavier than ChatGPT does, heavier than Claude. This is because Google owns YouTube and the transcripts are essentially first-party data for them. If your founder, your CEO, or your product expert has done any kind of video content in the last three years, and that content lives on YouTube with auto-generated or human-written transcripts, those transcripts are fair game for Gemini citations.
I had a B2B SaaS client land in Gemini's answer to "what's the standard contract structure for a SaaS reseller program" because their founder did a 45-minute interview on a niche industry podcast in 2023. The podcast was uploaded to YouTube. Gemini pulled three sentences from minute 17 of the transcript. The citation was the YouTube video — but anyone who clicked through learned about the company.
Tactical implications:
- Caption every YouTube video you put up. Auto-captions are fine for a baseline, but human-edited captions are noticeably cleaner and more quotable.
- Pitch yourself to industry podcasts and ask them to upload to YouTube if they don't already. Many podcasters will, especially if you mention it during recording.
- Consider repurposing your best blog content as 5–10 minute talking-head videos. Gemini cares about the transcript, not the production quality. A simple webcam recording with a clean transcript can outrank a fancy produced piece with no captions.
- Use the YouTube description field for a tight 2–3 sentence summary plus a few key timestamps. This gives Gemini metadata it can use to decide whether to pull from the transcript.
This is one of those leverage points where the work is unglamorous, the payoff is real, and almost no other "AI SEO" piece will mention it.
Schema for Gemini specifically
Gemini, being a Google product, respects Google's full structured-data ecosystem. The most useful types for AI visibility:
Article / BlogPosting. For anything that's a written piece, mark it up. Include author (with a Person sub-object including credentials), datePublished, dateModified, publisher. Gemini specifically uses the publication and modification dates to decide freshness, which matters more for Gemini than for ChatGPT.
FAQPage. Same as it is for ChatGPT — Gemini eats this format. Stick to genuinely useful Q&A pairs that match your visible markup exactly. Google will penalize discrepancies and you'll lose the rich result.
HowTo. If your page is a step-by-step, mark it up properly. Each step gets a HowToStep with a name and a text. Gemini will lift the steps verbatim if it decides to cite you.
Organization and LocalBusiness. If you're a local business, the full LocalBusiness markup with hours, address, payment methods, service area, and aggregateRating from a real review source — this is bread and butter. Local AI queries pull from this heavily.
Person (for author bios). If you publish content under named authors, give each one a real Person schema with sameAs links to their LinkedIn, their X account, their personal site. Gemini uses these to assess author authority, which is a real signal.
What I don't bother with for Gemini specifically: Speakable (intended for voice readers, niche), Quiz (irrelevant), SoftwareApplication unless you're literally pushing an app.
The freshness angle Gemini cares about
Gemini weights content freshness more aggressively than ChatGPT does. This is because Google has decades of experience with freshness signals (QDF — "query deserves freshness" — has been a thing since 2007), and Gemini inherits that algorithmic instinct.
What this means tactically:
- Update your evergreen content quarterly. Not just a date change — a real "what changed since last review" pass.
- Use
dateModifiedin your schema and update it when you actually modify the piece. - Add a visible "Last updated" line on the page. Gemini reads this. Humans appreciate it.
- If you have time-sensitive content (pricing, policies, year-specific data), put the year in the title. Pages titled "Best CRMs for agencies (2026)" outperform "Best CRMs for agencies" in Gemini answers for time-sensitive queries by a meaningful margin.
The "Gemini but not ChatGPT" story
I promised this story earlier. The B2B SaaS client showing up in Gemini but not ChatGPT — here's what was going on.
Their robots.txt was permissive for Googlebot and Google-Extended, so Gemini could happily ingest their content. But their previous developer had added a global block on GPTBot "for security" two years prior, never told anyone, and the block was buried in a Cloudflare rule set, not in robots.txt. So OpenAI's training crawler couldn't read the site at all. OAI-SearchBot was technically allowed in robots.txt, but Cloudflare's bot management was blocking it at the edge.
The fix took me eleven minutes. Turn off the Cloudflare AI-bot block. Confirm by curling the site with a fake GPTBot user-agent and getting back real HTML instead of an interstitial. Wait two weeks. Run manual ChatGPT queries. They started showing up.
The lesson: each model has its own consent and its own access. Just because you're winning in one doesn't mean the others can even read you. And the configuration is almost always somewhere subtle that the average site owner has never looked.
AI Overviews are a different game
AI Overviews specifically deserve their own section because they affect SEO in a way the Gemini app doesn't. When Google decides to show an AI Overview at the top of a search result, the traditional blue links get pushed down. Studies have shown traffic drops of 30–60% for some queries when an AI Overview appears, even when the overview cites you.
This creates a strategic question: do you want to be in the AI Overview, or do you want to push the AI Overview to not appear?
You can't directly control whether Google shows an AI Overview. But you can influence it:
- For queries where you want clicks more than citations: write content that doesn't pre-answer the question in a snippet-friendly way. Lead with context, then answer. AI Overviews tend to skip queries where no source provides a clean snippet.
- For queries where you want to be the cited source: do the opposite. Lead with the answer, mark it up cleanly, make it the easiest source to pull.
- For queries where the AI Overview is unavoidable: accept it and make sure you are at least one of the cited sources. Some click-through is better than none.
This is a real tradeoff and I don't have a one-size-fits-all answer. For most local service businesses I work with, citations help — the click rate is lower but the user is much higher intent. For content publishers monetizing through ads or sponsored posts, AI Overviews are an existential threat and the strategy needs to be different.
Tracking Gemini visibility
Google still hasn't shipped a real "Gemini Search Console" as of this writing, but here's what I use:
Google Search Console performance report. Filter by "Search appearance: AI Overviews" — yes, this is a real filter now in GSC, added quietly in 2024. It shows impressions and clicks specifically from AI Overview placements. If your site is cited in any AI Overview, you'll see it here.
Server logs. Look for hits from Google-Extended user agents. These are training and AI-answer crawls.
Manual queries in the Gemini app. Same as the manual-ChatGPT-queries advice in the ChatGPT piece. Once a week, run 10 queries you should rank for. Note the citations.
Manual queries in Google Search with AI Overviews on. Different surface, different results, different click economics. Track both.
A 30-day Gemini plan
If you give me a site tomorrow and tell me to maximize Gemini visibility in 30 days, this is exactly the order I'd run it in.
Week one. Confirm Google-Extended is allowed in robots.txt. Confirm no CDN or firewall is silently blocking it. Submit your sitemap to GSC if you haven't. Make sure every important page is indexed. Fix any indexing issues that aren't already fixed.
Week two. Add Article or BlogPosting schema with full author and date metadata to every blog post. Add FAQPage schema to your highest-intent commercial pages. If you're a local business, fully populate your LocalBusiness markup. Add author bios with real credentials and Person schema.
Week three. Caption every YouTube video your business has uploaded. If you don't have any, record a 5–10 minute talking-head version of your top blog post and put it up with a clean transcript. Update your three most important blog posts with fresh information and bump the dateModified.
Week four. Audit AI Overview performance in GSC. For queries where you're not cited, check who is — that's your gap. Rewrite content for the top three gaps with answer-first structure. Run 10 manual Gemini queries and note citations.
That's the realistic 30-day plan. None of this is sexy. All of it compounds.
What I'd skip
Quick list of "Gemini SEO" advice I see floating around that I would not bother with:
- "Optimize for conversational keywords." Long-tail keywords are great but you don't need a special Gemini strategy for them. Just write naturally.
- "Add a chatbot to your site so Gemini learns from it." No it won't, the bot interface isn't crawlable in the way you'd hope.
- "Use Google's AI tools to generate your content so Gemini will favor it." There is zero evidence Google does this, and Google has explicitly said they don't.
- "Pay for inclusion in a 'Gemini-optimized directory.'" These are scams. There is no such thing.
The takeaway
Gemini is a Google product, and the work to win it is mostly the work to win at Google — clean indexing, structured data, real authority signals, fresh content — with a few specific additions: the Google-Extended consent flag, full advantage of YouTube transcripts, and a careful read on whether you want to be in AI Overviews for a given query.
If you're already strong in Google Search, you can be cited in Gemini within weeks. If you're not, fix Google first — the rest follows.
And go check your robots.txt for Google-Extended. Right now. If it's not there, add it.
This piece will be kept up to date as Google ships new AI features and Gemini behavior changes. Last updated May 21, 2026.