Why This Guide Exists: AI Erased 101,820 Visits From Our Website Last Year

That’s a 31% drop in organic search traffic in 12 months. Real numbers, straight from our Google Analytics.

Replacing those clicks with Google Ads would cost roughly $535,000 a year at average CPC. B2B keywords run higher.

We run an answer engine optimization practice. We watch AI reshape search for clients every day. Our own data tells the same story.

The questions that used to bring people to websites now get answered inside ChatGPT, Perplexity, and Google’s AI Overviews. By 2026, fewer than a third of Google searches end in a click (SparkToro). That traffic is not coming back.

Two moves matter now. Become the source AI cites. Build for the smaller, smarter audience that still shows up. This guide covers both.

Adapting to AI-Driven Search with Generative Engine Optimization (GEO)

Search didn’t die. It moved. Your traffic went with it.

Buyers stopped typing keywords into Google. They ask ChatGPT and Perplexity full questions and get full answers. Google now answers queries right on the results page. No click. No visit. No lead.

This is not a reason to abandon SEO. Answer engines pull their sources from search results. If you don’t rank, you don’t exist to them. SEO gets you in the room. GEO gets you quoted. The two disciplines overlap but they are not the same, and we break down how optimizing content for LLMs differs from SEO in a separate guide.

What This Guide Covers

• How LLMs and answer engines actually find and pick content
• Why ranking in Google and Bing still decides whether AI cites you
• 12 content techniques that increase your odds of being quoted
• Schema markup and technical signals AI systems read
• How to track your brand inside AI answers

Understanding AI-Search

Where do AI systems get content from?

Two places. Get this distinction right and the rest of GEO makes sense.

1. Training data. Models learn from books, code, and huge swaths of the web, plus licensed sources like Reddit and major news publishers. This shapes what AI knows about your brand. It moves slowly. You influence it by being mentioned across the web, not just on your own site.

2. Live retrieval. Ask ChatGPT, Perplexity, or Google’s AI Mode a question and they search the web in real time and quote what they find. Most citations come from here. It moves fast. You influence it with pages that rank, load clean, and answer the question near the top.

Training data decides whether AI knows you. Retrieval decides whether AI cites you. Most companies obsess over one and ignore the other.

How do AI answer engines find your content?

Two ways, and both matter.

They crawl. OpenAI, Anthropic, Perplexity, and Google all run their own bots: GPTBot, ClaudeBot, PerplexityBot, Google-Extended. If your robots.txt blocks them, you’ve opted out of AI visibility. Check this before you spend a dollar on GEO. We’ve seen companies pay for AI visibility consulting while their own site blocked respectful AI crawlers.

They retrieve on demand. When someone asks a question, the answer engine runs a live search, pulls the most promising pages, and quotes what answers the question. ChatGPT leans on Bing plus its own index. Perplexity built its own. Google’s AI Mode uses, well, Google.

Retrieval rewards pages that are easy to lift answers from:

• Clear structure with headings that match real questions
• A direct answer near the top, not buried under a 400-word warmup
• Clean HTML that’s easy to quote

Long pages are fine. Unscannable pages are not. If a model can’t find your answer in seconds, it quotes someone else.

Understanding LLM Behavior

Large Language Models (LLMs) are AI systems trained on vast text datasets to generate human-like language. Here’s the multi-step process they run when they search the web on your behalf:

The Changing Search Landscape

People search more and click less. Google’s AI Mode passed one billion monthly users within a year of launch, and Google reports total queries at an all-time high. Sounds like good news for websites. It isn’t. The queries grew. The clicks didn’t.

AI Overviews and AI Mode resolve the question on Google’s page, and ChatGPT and Perplexity resolve it before Google is even involved. Google

That’s the math behind our own 31% drop. More questions asked. Fewer answered by a visit to anyone’s website.

Paul Baier, CEO, GAI Insights

By mid-2025, Baier declared the bloodbath “well underway,” pointing to traffic at major publishers dropping by half while ChatGPT climbed into the world’s top five websites. Our analytics agree. He undershot.

Optimizing Content for Inclusion in LLMs

1. Create In-Depth, Comprehensive Content

LLMs prioritize thorough, well-structured information. Implement these tactics:

• Cover All Angles: Explore your topic exhaustively, including related subtopics. Showcase your deep knowledge of the topic. This comprehensive approach helps LLMs understand the full context.
  • Include insights, analysis, and information that clearly establish your authority in the field.
  • Aim for substantial word counts (e.g., 1500+ words for key pages) when appropriate to showcase comprehensive knowledge.
• Use Examples and Case Studies: Incorporate real-world examples and up-to-date data. This enriches your content and helps LLMs make accurate connections between concepts.
• Define Key Terms: Explain technical terms and jargon clearly. This ensures both human readers and AI systems can fully comprehend your content. Use semantically-related terms and concepts.
• Organize Logically: Use descriptive headings and subheadings to structure your content. A well-organized article is more likely to be accurately processed and categorized by LLMs.

2. Focus on creating comprehensive FAQ pages

FAQ pages provide direct answers to common questions, serving as a valuable resource for both users and LLMs. They align well with natural language queries, including voice searches, and offer a centralized hub for detailed information. Use keyword research tool, “Answer the Public” and Google’s “People Also Ask” feature to identify popular questions in your niche. Answer these questions concisely, using 1-2 sentences per answer.

3. Write the Way Buyers Ask

LLMs are trained on natural language. Match it.
Use full questions, not keyword fragments. “How do I choose enterprise software?” beats “enterprise software solutions.” Pick strong verbs: “buy,” not “make a purchase.” Use contractions. Read it aloud. If it sounds like a brochure, rewrite it.

4. Optimize for Featured Snippets

Featured snippets show up less often now. AI Overviews have taken over many of the question-style queries that used to trigger them, and snippet visibility has dropped hard since 2024. Source: Keywords Everywhere

So why is this technique still here? Because the snippet became the audition.  Pages that win snippets are the pages AI Overviews cite. The same patterns win both: clear question framing, a direct first-sentence answer, formatted lists and tables.

5. Enhance Content Structure and Presentation

While LLMs primarily process text, well-structured content is more likely to be accurately interpreted:

• Use bullet points and numbered lists for easy scanning and information extraction.
• Break up long paragraphs into shorter, more digestible chunks.
• Include descriptive captions for any images or videos to provide context for LLMs.
• Use consistent formatting and structure across your content to aid in pattern recognition.

6. Use Diverse Content Formats

Incorporate a mix of text, video, and images in your content. This diversity not only caters to different user preferences but also provides multiple ways for AI systems to understand and contextualize your information. Ensure each format complements the others, reinforcing your key messages across different mediums.

7. Enhancing Content Credibility, Authority, and AI Visibility

Establishing your Authority in AI Search results:
Research: Conduct and publish original research or data analysis to position yourself as a primary source.

Author Bios: Clearly state relevant qualifications in author bios to reinforce your expertise to both readers and AI systems.

Key strategies to implement:

• Credible Sources: Include expert quotes, case studies, and credible citations to improve visibility, add real-world context, and enhance trustworthiness.
• Fact-based information: Integrate relevant statistics and data to support the content, making it more informative and reliable.
• Readable and Clear Language: Optimize the fluency and simplicity of the text to enhance readability and understanding.

8. Craft Persuasive Content

Modify your content to be more persuasive by making authoritative claims backed by evidence. Use confident language, provide compelling arguments, and clearly articulate the value proposition of your ideas or products. This approach not only convinces human readers but also signals to AI systems the strength and relevance of your content.

9. Enhancing Cross-linking for AI Search

To improve AI search visibility, create a well-connected content structure. Organize topics into clusters with pillar pages, use descriptive anchor text for internal links, and include related article sections. This approach helps AI systems better understand your content’s context and relationships, potentially boosting its relevance in search results.

10. Monitor Performance Across Both Traditional and AI-Powered Search Results

Rank trackers can’t see inside ChatGPT. You need an AI visibility platform. We use REVERE to track brand mentions, citations, and competitor share across the major AI engines.

Measure three things: are you mentioned, are you cited, and who wins the answer when you don’t.

Schema Markup and Technical SEO: Boosting AI Comprehension

1. Use Structured Data

Structured data gives search engines and AI systems machine-readable context. Your page says what it is instead of making the machine guess. Mark up the things machines get wrong: products and services, events and dates, organization details, and article type.

FAQ Schema Still Matters. The reason changed.

Google killed FAQ rich results for most sites in 2023. The old pitch, more SERP real estate, is dead.

The new pitch: FAQ schema hands AI systems your questions and answers as labeled pairs. No parsing needed. That’s the format answer engines want.

Write real buyer questions as headings. Answer in one or two sentences. That’s how people prompt ChatGPT.

2. Title Tag Optimization: Balancing SEO and LLM Considerations

The 2024 Google Search API leak confirmed title tags matter for rankings. LLMs read them too, when crawling and when citing.

Write concise, keyword-rich titles that say what the page is.

Preview how your meta title and description tags will appear.

3. Optimize for Wide Distribution – go beyond Google

LLMs often recognize content that appears across multiple reputable websites. Aim to have your content cited, quoted, or referenced on various platforms to increase its visibility to AI systems. For example, increase your PR spend to get to get your message more widely distributed.

Danny Sullivan, Google Search : ‘To succeed with Google Search, think beyond it.’

4. Encourage User Generated Content

AI engines cite Reddit threads, reviews, and forums constantly, often above brand websites. When buyers ask “is [your product] any good,” the answer comes from what your customers wrote, not what you wrote. Earn reviews on the platforms your buyers trust. Show up in the communities where your category gets discussed.

6. Optimize Video Content

YouTube is now the most-cited domain in Google’s AI Overviews, with citations up 34% in six months.

LLMs don’t literally watch footage; they read the text YouTube provides about the video: titles, descriptions, tags, chapters, captions/transcripts, and comments. Source: aHrefs

These 3 steps increase discoverability in both traditional search and AI-generated recommendations.

1. Make those elements accurate and answer-focused so models can return your videos in responses.
2. Use descriptive, question-based titles and chapter names. Upload or clean up transcripts instead of relying only on auto-captions.
3. Add clear descriptions, targeted tags, and video sitemaps. Encourage likes, comments, and shares.

Guardrail: Don’t outsource quality to AI

Some ad networks have dropped publishers that flooded their sites with low-quality AI content.The problem isn’t using AI. It’s publishing repetitive, unedited content at scale.

When the marginal cost of creating new and customized content plummets, which it will as LLMs become even more capable, then content will explode, even more than it has today. The traditional web will be overwhelmed with AI generated content. The time to plan for this tsunami is now.

John Sviokla, Executive Fellow @ Harvard Business School | D.B.A., GAI Insights Co-Founder

These strategies enhance your content’s compatibility with LLMs. You’ll improve your chances of appearing in AI search results. This approach not only caters to AI systems but also creates more user-friendly content for human readers, potentially increasing engagement and reach.

Q&A from our Expert Panel

We’ve covered how AI is changing search. We’ve shared strategies and actionable steps to future-proof your presence. However, as we discussed, the new AI-driven landscape is nuanced and continuing to evolve. For that reason we decided to reach out to some experts in the field.

In July, I hosted an AI Learning Lab panel with two experts: Alden Do Rosario (CEO, CustomGPT) and Julio Barros (President, E-String).

#1 – Can LLMs be influenced to reflect personal views or content for other people?

It’s a common belief that LLMs can be manipulated to consistently favor a specific viewpoint. We asked our panelists weigh in: timestamp :2:54 min

• Yes, through data sources like Bing search and Common Crawl
• Limited control, depends on fixed knowledge and search tools
• Individual users can’t significantly influence LLMs
• Group behavior might influence LLMs over time

#2 – Can repeated user interactions with an LLM shape its responses over time, potentially biasing it towards specific perspectives or information sources?

• Individual user interactions do not currently shape LLM responses over time
• LLMs don’t monitor user behavior like page visits or time spent on results
• While an individual can’t bias the model, group behavior at scale may influence LLMs over the long term
• Companies like OpenAI likely use aggregate user data to improve their models
• The potential for collective user influence raises ethical questions about LLM training and output

#3 – What influences an LLM’s decision to cite sources, and how does this reflect its understanding and information retrieval? <30 min>

• LLMs typically cite sources when they’ve used external tools like Bing search or partnerships with news/Reddit
• Without external tools, LLMs rely on their training data and don’t cite specific sources
• The decision to cite is often influenced by the system prompt given to the LLM
• ChatGPT is transparent about when it’s using external search, showing this to users
• Source citation reflects the LLM’s certainty about where specific information came from

Your next steps

Here’s your action plan to thrive in the age of AI search:

1. Revamp Your Content Strategy: Create comprehensive, well-structured content that addresses user queries in-depth. Remember, AI loves context and thoroughness.
2. Optimize for Conversational Queries: Match the way people ask questions. This serves voice search and chatbots.
3. Leverage Structured Data: Implement schema markup, especially FAQ schema, to help AI systems better understand and categorize your content. Don’t forget to try our free FAQ schema generator tool to simplify this process!
4. Monitor and Adapt: Regularly check how your content performs in AI-driven search results. Use tools like ChatHub to compare responses across different AI models and adjust your strategy accordingly.
5. Balance AI and Human Touch: While optimizing for AI, never lose sight of your human audience. Blend data-driven insights with creativity and genuine value.
6. Stay Informed: The world of AI search is evolving rapidly. Subscribe to our newsletter/blog for the latest updates and strategies in AI search optimization.

June 10/2026