Generative Engine Optimization Guide

Bottom Line

Generative engine optimization (GEO) means making your content easy for ChatGPT, Perplexity, and Google AI Overviews to find, trust, and cite. Any honest generative engine optimization guide starts with Google's May 2026 position: GEO is still SEO. The work is familiar. Write clear answers, cite real sources, keep pages crawlable, and track citations over months.

Key Takeaways

• Google says optimizing for AI search is still SEO, so keep your fundamentals.
• Adding citations, statistics, and quotes raised AI visibility up to 40% in testing.
• LLM referral traffic is still under 0.2% of e-commerce traffic in one study.
• AI answers change almost every run, so track share of voice, not rankings.
• Short-term GEO tricks tend to work, spread, and then get patched.

Organic click-through rates on Google queries with an AI Overview fell 61% between June 2024 and September 2025. They dropped from 1.76% to 0.61%, according to Seer Interactive's study of 3,119 terms and 25.1 million impressions. That number explains why "GEO" went from niche term to boardroom topic in about a year.

If you run a Shopify store or a small service business, you don't have time for a new discipline. The good news is you probably don't need one. This generative engine optimization guide covers what actually changed, what didn't, and where the hype outruns the data.

GEO Is Still SEO, According to Google

Generative engine optimization is the work of getting your content cited inside AI-generated answers. Google's first official guide to AI search, released May 15, 2026, calls it part of SEO. The core skills stay the same: useful content, clear structure, and crawlable pages. What changes is how you format answers and measure success.

The many names for the same idea

You'll see a pile of acronyms for this work:

• GEO: generative engine optimization
• AEO: answer engine optimization
• LLMO: large language model optimization
• AI SEO: the catch-all term

They describe the same goal. You want an AI tool to quote your page when someone asks a question in your niche. Don't waste time picking the "right" label.

What Google said in May 2026

Google's guide was blunt. It says "optimizing for generative AI search is optimizing for the search experience, and thus still SEO" (Google via Semrush/Wikipedia, 2026). Most GEO guides were published before that statement. Many of them treat GEO as a separate playbook with its own secret rules.

Google's framing is the more useful one. AI Overviews and AI Mode pull from Google's own index. If your page isn't indexed and trusted, it won't get cited.

Where GEO and SEO actually differ

The overlap is large, but a few things do shift.

| Dimension | Classic SEO | GEO | |---|---|---| | Unit of success | A ranking position | A citation or mention in an answer | | Input you research | Keywords | Prompts and full questions | | What gets extracted | The whole page | One passage or section | | Measurement | Stable rank tracking | Share of voice across many runs | | Main traffic source | Google organic clicks | Mostly brand exposure, some referrals |

AI Engines Are Changing Where Clicks Go

AI answers now sit above the classic results, and they absorb clicks. Pew, Seer Interactive, and Ahrefs all found sharp click-through drops when an AI summary appears. LLM tools send very little referral traffic today, though. You need to plan for both facts at once.

The click squeeze is real

The data points in the same direction:

• Pew Research Center (2025): Users clicked a traditional result in 8% of visits with an AI summary, versus 15% without one.
• Ahrefs (2026): AI Overviews now mean a 58% lower average click-through rate for top-ranking pages. That's up from 34.5% just eight months earlier.
• Tech Insider (2026): At Google I/O on May 19, 2026, Google said AI Mode had crossed 1 billion monthly users. It now runs on Gemini 3.5 Flash by default.

Ranking first doesn't guarantee the click it once did. Being the cited source inside the answer is the new front row.

LLM referrals are still small

Here's the part most GEO pitches skip. A 2025 study in Marketing Science (INFORMS) looked at 973 e-commerce sites. LLM traffic was less than 0.2% of all traffic. That's about 200 times smaller than Google organic search.

ChatGPT itself is huge. OpenAI reported 900 million weekly active users as of February 2026, up from 400 million a year earlier. But big usage hasn't turned into big referral volume for most sites. People read the answer and move on.

For a D2C store, Google organic still pays the bills. GEO is a hedge and a brand play, not a replacement.

Crawlers take more than they send

Cloudflare's July 2025 data shows how lopsided the trade is. Google crawled about 5.4 pages for every visitor it referred. OpenAI crawled roughly 1,091 pages per referred visitor. Anthropic crawled about 38,066.

This ratio matters for one decision. Blocking AI crawlers protects your content but removes you from their answers. Most small businesses gain more from visibility than they lose to scraping. Publishers with paywalled content may decide differently.

This Generative Engine Optimization Guide Starts With Citable Content

Citable content is content that AI engines can easily extract and easily trust. A Princeton-led study by Aggarwal et al. (KDD 2024) found five tactics that raised visibility in AI answers by up to 40%. The common thread is evidence: named sources, real numbers, and direct quotes placed right next to the claims they support.

The five tactics from the Princeton study

Aggarwal et al. presented the GEO study at KDD 2024. The researchers tested methods across a 10,000-query benchmark. These five tactics boosted citation rates:

1. Cite sources. Link to the primary study, not a blog that summarized the study.
2. Add quotations. Quote a named expert with the expert's role.
3. Add statistics. Give the number, the source, and the year.
4. Optimize fluency. Write clean, readable sentences.
5. Use an authoritative voice. State facts plainly, without hedging every line.

Here's a quick before and after. "Many shoppers abandon carts" is weak. "Shoppers clicked a traditional result in only 8% of visits with an AI summary (Pew Research Center, 2025)" is citable. The first sentence is an opinion. The second sentence is a fact a model can lift and attribute.

Answer blocks under every heading

AI engines pull one section at a time, so each section has to stand alone. Put a 40 to 60 word answer directly under each heading. Then expand the answer with detail, examples, and data.

Start from real prompts, not just keywords. A keyword is "shopify seo apps." A prompt is "what's the cheapest way to get blog traffic for my Shopify store without hiring someone." Write headings that match the second kind of question. Add an FAQ section for the follow-up questions people actually ask.

CocoSEO builds articles this way by default. Every CocoSEO draft includes researched, cited sources and answer-first sections. You review the draft in about 60 seconds and publish straight to WordPress, Shopify, or Ghost.

The 2023 benchmark caveat

The Princeton benchmark dates from 2023. ChatGPT, Perplexity, and Google's AI features have changed substantially since 2023. The 40% figure may not carry over to today's engines.

The underlying logic of the Princeton study still holds, though. Models that cite sources prefer passages with clear, attributable facts. Treat the study as directional, not a guarantee.

Technical Access Decides Whether You Get Read

Technical access is whether AI crawlers can reach and parse your pages, and AI engines can only cite pages they can reach. Good technical access means allowing crawlers like GPTBot, ClaudeBot, and PerplexityBot, serving content in plain HTML, and adding structured data like FAQPage schema. Newer ideas like llms.txt get attention, but Google says its search systems don't use llms.txt.

Let the right crawlers in

Check your robots.txt file. The main AI user agents include:

• GPTBot and OAI-SearchBot: OpenAI's crawlers for training and search
• ClaudeBot: Anthropic's crawler
• PerplexityBot: Perplexity's crawler
• Google-Extended: Controls Gemini training, separate from regular Search indexing

Some security plugins and CDN settings block these AI crawlers by default. Cloudflare added AI crawler blocking options in 2025. Test your site before assuming you're visible to AI engines.

Serve content in plain HTML

Many AI crawlers don't run JavaScript the way Googlebot does. If your product details or article text load only after scripts run, an AI crawler may see a blank page. Shopify themes and headless builds are common trouble spots.

A simple test is to view your page source in the browser. If you can't find your main text in the page source, an AI crawler may miss that text too.

Schema and llms.txt in perspective

FAQPage and Product schema help machines understand what a page is about. FAQPage and Product schema are cheap to add and support classic rich results too. A detailed walkthrough is coming in our guide to FAQ schema for small business sites.

llms.txt is a proposed file that lists your key pages for language models. Google says llms.txt isn't used by Google's Search systems (SEJ, 2026). Lighthouse lists llms.txt only as optional, in an experimental category. Add an llms.txt file if it takes five minutes. Don't expect llms.txt to move results.

Measuring GEO Takes a Different Mindset

GEO measurement is the practice of tracking how often AI engines cite your brand across repeated runs of the same prompts. Unlike a Google ranking, AI answers change from one run to the next, so a single check means little. The useful approach is to measure citation rate and share of voice, then watch that share move over weeks.

AI answers change almost every run

Rand Fishkin of SparkToro tested 2,961 prompts. Fewer than 1 in 100 runs produced the same list of brands. Fewer than 1 in 1,000 runs produced the same list in the same order.

So "we rank #2 in ChatGPT" is close to meaningless. You might be #2 now and missing an hour later. Any tool selling fixed AI rankings is overstating what that tool can measure.

Track share of voice over time

Better GEO metrics look like this:

• Citation rate: How often your domain appears across many runs of a prompt
• Share of voice: Your mentions versus competitors on the same prompt set
• Referral sessions: Visits from chatgpt.com, perplexity.ai, and similar sources in your analytics
• Search Console data: Impressions and clicks, which include AI Overview appearances

CocoSEO tracks citations across ChatGPT, Perplexity, and Google AI Overviews next to Search Console data. You get the citation data in a weekly digest. We'll cover setup in more depth in a post on how to track AI citations.

Treat case studies and spikes carefully

Some GEO results look strong. Acceleration Partners reported that a financial services brand doubled its AI share of voice. The financial services brand also raised LLM citations by nearly 60% (2026). Neur Digital and Otterly reported 8x more citations in 12 months for a medical device client (2025). Neur Digital and Otterly got that result by mapping competitor-heavy prompts to specific pages.

Note the timeline of the medical device case: 12 months. Also note that vendor numbers like +340% or +447% are self-reported claims.

Referral traffic swings hard too. Similarweb found BBC referrals from ChatGPT up 188.9% since January 2025, about 118,000 visits. Profound reported ChatGPT referral traffic to websites down 52% since July 21, 2025. One model update can erase a referral trend.

Shortcuts in GEO Get Patched Fast

A GEO shortcut is a tactic that games AI models instead of earning citations, such as self-serving listicles or prompt injection. GEO shortcuts follow the same arc old SEO tricks did: they work briefly, spread across forums, and then get patched. Lily Ray of Amsive and Algorythmic describes this cycle directly. Sustainable gains come from content that deserves citation.

The work, spread, patch cycle

Lily Ray argues that short-term GEO tactics usually work, then spread, then get patched (2026). The work-spread-patch cycle isn't new. Keyword stuffing and private link networks went through the same cycle in Google's early years.

The work-spread-patch pattern also costs you twice. You lose the gains, and you may carry a trust penalty after.

Two tactics to skip

Avoid these two GEO tactics, even if a competitor seems to get away with them:

1. Self-serving "best of" listicles. Ranking your own product #1 in "best tools" posts is easy to spot. AI engines and Google are both cracking down on self-serving listicles.
2. Prompt injection. Hiding text that tells an AI to recommend your brand is manipulation. Prompt injection risks removal from results entirely.

The boring path compounds. Clear answers, real sources, and steady publishing build visibility that survives model updates.

FAQ

Q: Is generative engine optimization different from SEO? Mostly no. Google's May 2026 guide says optimizing for generative AI search is still SEO. The differences are in formatting answers for extraction and measuring citations instead of rankings.

Q: How long does GEO take to show results? Expect months, not weeks. The Neur Digital and Otterly medical device case reported 8x more citations over 12 months. Because AI answers vary so much, you need weeks of repeated checks to see a real trend.

Q: Should I block AI crawlers like GPTBot? Most small businesses shouldn't. Blocking protects content but removes you from AI answers. Cloudflare's 2025 data shows AI crawlers take far more than they send. Weigh that against the brand exposure you'd lose.

Q: Does llms.txt help me get cited? There's little evidence that it does. Google says llms.txt isn't used by its Search systems. Lighthouse treats it only as an optional, experimental check.

Q: Will GEO replace my Google traffic? Not soon. A 2025 Marketing Science study of 973 e-commerce sites found LLM traffic under 0.2% of total traffic. Keep Google organic as your main channel and treat GEO as a growing extra.

Your Next Step

The next step in GEO is a single editing pass on your three most important pages, starting today. Add a 40 to 60 word answer under each main heading. Then add one cited statistic with a named source and year. That single editing pass applies the core of this generative engine optimization guide in under an hour.

Further reading

• "GEO: Generative Engine Optimization" by Aggarwal et al. (KDD 2024): The original Princeton-led study behind the five citation tactics.
• Google Search Central's May 2026 guide to AI search: Google's own explanation of why GEO is still SEO.
• SparkToro's AI answer consistency research by Rand Fishkin: The clearest data on why AI "rankings" are unstable.
• Seer Interactive's AI Overview CTR study (2025): The 3,119-term dataset behind the 61% click-through drop.
• Cloudflare's 2025 crawl-to-referral analysis: Useful context before you decide whether to allow or block AI crawlers.

Sources

1. AIO Impact on Google CTR: September 2025 Update — Seer Interactive, 2025. Supports: Organic CTR on queries with an AI Overview fell 61% between June 2024 and September 2025.
2. Google users are less likely to click on links when an AI summary appears in the results — Pew Research Center, 2025. Supports: Users clicked a traditional result in 8% of visits with an AI summary, compared with 15% without one.
3. AI Overviews Tracker (includes updated CTR study findings) — Ahrefs, 2026. Supports: AI Overviews now correlate with a 58% lower CTR for top-ranking pages, up from 34.5% in Ahrefs' earlier study.
4. Frontiers: ChatGPT Referrals to E-Commerce Websites: How Do LLMs Compare Against Traditional Channels? — Marketing Science (INFORMS), 2025. Supports: The 973-site e-commerce study finding LLM traffic under 0.2% of total traffic, about 200 times smaller than Google organic.
5. AI crawling data reveals massive imbalance in training versus referral patterns — PPC Land (reporting Cloudflare data), 2025. Supports: In July 2025 OpenAI crawled about 1,091 pages per referred visit and Anthropic about 38,000. The source does not verify Google's 5.4 figure.
6. GEO: Generative Engine Optimization — Aggarwal et al., KDD 2024 (arXiv). Supports: The Princeton-led study showing GEO methods can raise visibility in generative engine answers by up to 40%, tested on a 10,000-query benchmark.
7. NEW Research: AIs are highly inconsistent when recommending brands or products — SparkToro, 2026. Supports: Rand Fishkin's finding that fewer than 1 in 100 runs return the same list of brands.
8. Google publishes guide to optimizing for generative AI search — Semrush, 2026. Supports: Google released its official guide on optimizing for generative AI search on May 15, 2026, and framed GEO as still SEO.