Generative AI SEO: What Changes and What Stays the Same

Most advice on generative AI and SEO falls into two camps. One says you need an entirely new playbook for AI search, with new file formats, new markup and content chopped into machine-sized pieces. The other says nothing has changed. Both are wrong in ways that cost money, and the evidence sits somewhere in between.

Key Takeaways

• Google states that optimizing for generative AI search is optimizing for the search experience, and therefore still SEO. Google also says no special schema markup is required.
• AI summaries are associated with fewer clicks. Pew Research Center measured 8% of visits with a click on a traditional result when an AI summary appeared, against 15% without one.
• Generative AI is a legitimate tool for research, structure and first drafts. Google treats mass-produced pages that add no value as scaled content abuse, whoever or whatever wrote them.
• Published research on getting cited by AI engines is promising but early. Treat each tactic as an experiment you measure, not as a rule.
• Every AI output needs a human check before it goes live, including titles, meta descriptions, structured data and alt text.

Why does generative AI change your SEO results?

Generative AI changes SEO results because search engines now write answers themselves, which can satisfy a query without sending a visit. For a business, the practical effect is fewer clicks per impression on informational queries, and a second place to be seen: inside the answer.

That affects the numbers a board asks about. Traffic from informational pages can fall while rankings hold steady, and a report that only tracks sessions will read that as failure when it may be a change in how results are displayed. Your reporting needs to separate impressions, clicks and, where you can measure it, mentions inside AI answers.

What is generative AI SEO?

Generative AI SEO is the work of making a website findable and citable in AI-generated search answers, and of using generative AI tools to research, draft and improve search content, with a human responsible for accuracy.

Clients use the term for both halves. Search engine optimization (SEO) is the discipline of earning visibility in search results. The glossary holds the shorter versions of both terms.

We cover both halves here because they pull in different directions. Using AI to produce content faster tempts teams to publish more, while being cited by AI systems rewards fewer, better, more checkable pages.

What does Google say about optimizing for AI search?

Google says optimizing for generative AI search is the same job as ordinary SEO. Its Search Central guidance, last updated in July 2026, states that “optimizing for generative AI search is optimizing for the search experience, and thus still SEO.”

The same guidance says structured data (markup that labels page content for machines) is not required for generative AI search and that no special schema is needed. It says Google Search does not use llms.txt files, and that you do not need to break content into tiny pieces or write in a particular style for AI. If a vendor sold you any of those as a requirement for Google, the vendor is ahead of the evidence. Schema still earns its place for other reasons, such as rich results in classic search, which our structured data checklist covers.

One caution. That guidance describes Google. ChatGPT, Perplexity and other assistants run their own systems and publish less about how they choose sources, so what is true for Google Search should not be assumed for them.

How much do AI summaries reduce clicks?

AI summaries are associated with fewer clicks on ordinary results, and two independent studies point the same way. The size of the effect depends on the query, the ranking position and the method, so the figures below are evidence of direction, not a forecast for your site.

Pew Research Center found that Google users clicked a traditional search result in 8% of visits that showed an AI summary, compared with 15% of visits that did not.

Pew analysed browsing activity from 900 US adults who agreed to share it, covering Google searches in March 2025. Source: Pew Research Center, 22 July 2025.

Ahrefs reported that an AI Overview on a results page correlates with a 58% lower average click-through rate for the top-ranking page.

The analysis covered 300,000 keywords, split into two groups of 150,000, comparing December 2023 with December 2025. Ahrefs describes the result as a correlation, not proof of cause. Source: Ahrefs, Ryan Law and Xibeijia Guan, 2026.

Researchers at NJIT, NTU and Indiana University Bloomington found that Google AI Overviews appeared for 51.5% of representative real-user queries.

Their data was collected on 7 and 8 December 2025 from US locations. The sources cited by classic results, AI Overviews and Gemini overlapped little, with an average Jaccard similarity (a 0 to 1 measure of shared items) below 0.2. Source: Grossman et al., SIGIR 2026.

Here is the chain most people assume: an AI summary appears, the reader gets the answer, the reader does not click, and traffic falls. The studies support the first and last links in aggregate. They do not prove the middle ones for your pages. Informational queries were already low-click before AI summaries, and those are also the queries most likely to trigger one, so part of the gap may come from the type of query rather than the summary. Pew observed 900 people, and the Ahrefs and Grossman datasets are snapshots of Google at one moment. Check your own Search Console data by query type before you rewrite a strategy around any of these numbers.

The Grossman finding has a practical edge. If classic results and AI answers draw on different sources, a page ranking third for a query is not automatically cited in the answer above it. In our own work we treat ranking and citation as two separate measurements, which is the idea behind our guide to measuring AI search visibility.

A traffic drop and a ranking drop are different problems, and a report that cannot tell them apart will send you to fix the wrong one.

Can you use generative AI to produce SEO content?

Yes, with conditions. Google says generative AI can help with researching a topic and adding structure to content, and that generating many pages without adding value for users may violate its spam policies on scaled content abuse. Google also tells site owners to fact-check AI output, including titles, meta descriptions, structured data and alt text.

That makes the useful question “which steps does the tool do well, and which steps need a person?” Our Human Algorithm is Wild Creek’s own framework for answering it: the tool produces the speed, and a person supplies the judgement and takes responsibility for the result. It is our working model, not an industry standard, and other practitioners draw the lines differently. The table shows where we draw them for SEO work. You can read the full model on the Human Algorithm page.

How should you prompt AI for SEO work?

Prompt engineering, the practice of writing instructions that get a language model to produce usable output, comes down to giving the model a role, your context and a defined output. Vague prompts return generic text. The examples below are starting points to adapt, and our prompt engineering guide goes further.

Ask a language model for search volumes and it will give you plausible numbers that no tool measured. Treat any volume or difficulty score that did not come from a keyword tool as unverified.

Do the research tactics for getting cited by AI work?

The main academic study suggests some tactics can raise visibility in AI-generated answers, but it tested a research benchmark, not live engines, so treat the results as hypotheses. Aggarwal et al. introduced Generative Engine Optimization, or GEO, in a paper at the 2024 ACM SIGKDD conference. They built GEO-bench, “a large-scale benchmark of diverse user queries across multiple domains”, and reported that GEO methods “can boost visibility by up to 40% in generative engine responses.”

Two limits apply. “Up to 40%” is a ceiling observed under test conditions, not an expected result, and the engines in the study have changed since 2024. We therefore run these tactics as experiments with a measured before and after, such as adding a cited statistic to one page and tracking whether the page is mentioned in answers over the following weeks. For the wider approach, see our pages on LLMO strategy and how to rank in AI Overviews.

An AI draft that nobody has fact-checked is a liability formatted like an article.

What should you do about generative AI SEO this week?

Start by finding out where AI summaries are costing you, then fix accuracy before you add new tactics. The checklist takes about a working day for a small site.

• Export your top 50 pages by impressions from Search Console and flag the ones with high impressions and a falling click-through rate.
• Check whether those flagged queries show an AI summary when you search them yourself, and note whether your brand is named in it.
• Read every AI-assisted page published in the last 12 months and verify each statistic, date and claim against its source.
• Audit what you have been sold as “AI SEO”, such as llms.txt files or special schema, and ask the vendor for evidence that it affects Google.
• Pick one page, add one verified, sourced statistic and one first-hand example, and record the date so you can compare mentions later.

If you work with an agency, ask which of these five they already do, and ask to see the output. Our AI SEO services page describes how we approach this, and human-AI collaboration in SEO explains the working split in more detail.

Where to start

Before you buy a new tool for AI search, spend one hour on the first checklist item. A list of pages with rising impressions and falling clicks tells you whether you have a visibility problem at all, and which pages are worth a rewrite. If the list is short, put the budget into fact-checking and improving the pages you already have.

Frequently Asked Questions

Will generative AI replace SEO?

No evidence we have reviewed shows generative AI replacing SEO, and Google itself describes optimizing for AI search as still SEO. A Gartner survey of US consumers released in January 2026 found only about one-third believe GenAI chatbots are as effective as search engines for learning new information. Gartner’s own advice was that visibility now means optimizing for both AI-driven answers and classic search results.

Does Google penalize AI-written content?

Google’s guidance does not treat the use of AI as a violation in itself. It targets scaled content abuse, meaning many pages generated without adding value for users. Whether a person or a model wrote a page matters less than whether the page is accurate and useful.

Do I need llms.txt or special schema for AI search?

For Google Search, no. Google states that structured data is not required for generative AI search and that Google Search does not use llms.txt files. Other AI assistants may behave differently, so test any such file as an experiment instead of a requirement.

How do I know if my brand appears in AI answers?

Search your 20 most important queries in the AI features you care about and record whether your brand or pages are mentioned, then repeat on a fixed schedule. Google also points site owners to a Search Console report on generative AI performance, and it advises caution with third-party tools that claim to show Google internal metrics.

Can small businesses compete in AI search?

Small businesses can compete on specificity. The Grossman study found classic results and AI answers draw on different sources, so a narrow page with first-hand detail may be cited even when a larger site outranks it. We cannot promise that outcome, and it needs to be tested for your market.

Sources & Further Reading

• Google Search Central, “Google Search’s Guidance on Generative AI Content on Your Website”, Google, 2026.
• Google Search Central, “Optimizing your website for generative AI features on Google Search”, Google, 2026.
• Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results”, 22 July 2025.
• Ryan Law and Xibeijia Guan, “AI Overviews Reduce Clicks” (update), Ahrefs, 2026.
• Riley Grossman, Songjiang Liu, Michael K. Chen, Mike Smith, Cristian Borcea and Yi Chen, “How Generative AI Disrupts Search: An Empirical Study of Google Search, Gemini, and AI Overviews”, SIGIR 2026 (arXiv:2604.27790).
• Pranjal Aggarwal et al., “GEO: Generative Engine Optimization”, 30th ACM SIGKDD Conference, 2024 (arXiv:2311.09735).
• Gartner, “Gartner Survey Finds Only One-Third of Consumers Say GenAI Rivals Search Engines”, press release, 20 January 2026.