AI and SEO in 2026: what GEO changes, what it doesn't, and what the best GEO agencies do

The short answer: AI search hasn't replaced SEO. It has added a second question on top of it. SEO asks "do we rank?". Generative engine optimization (GEO) asks "when someone asks an AI assistant, are we named, and how are we described?". The foundations are the same: pages that can be crawled, clear answers and real reputation. What changes is that third-party mentions matter more, the "result" is a sentence rather than a link, and you have to measure it differently. The best GEO agencies treat it that way: as SEO plus reputation work, measured in AI answers, not as a bag of new tricks.

This guide explains how AI search works, what actually changes, which popular tactics don't hold up, and how AI is changing the day-to-day work of SEO and GEO.

First, the terms

You'll see several names for roughly the same thing:

• GEO (generative engine optimization): improving how often and how well your brand appears in answers from generative AI, such as ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews. The term comes from a 2023 research paper by researchers from Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi.
• AEO (answer engine optimization): an older term for the same goal, from the days of featured snippets and voice assistants.
• AI SEO / LLM SEO: marketing shorthand, usually meaning either GEO or "using AI tools to do SEO". This article covers both.

A GEO agency is a company that does this work for you. A GEO tool is software that measures it. We compare the two in GEO agency vs GEO tool.

How AI search actually works

An AI assistant answers a question in one of two ways.

From memory. The model answers from what it learned during training. That reflects what the web said months or years ago, so you can't change it quickly.

By searching first. For anything current, or anything that sounds like a shopping or comparison question, most assistants run web searches, read a handful of pages and write an answer from them. This is where GEO has the most effect, because the pages it reads can change this week.

A few details from the platforms themselves:

• Google says AI Overviews and AI Mode may use "query fan-out": one question becomes several related searches across subtopics, so the answer can draw on a wider set of pages than a single search would.
• ChatGPT says its search uses "third-party search providers, as well as content provided directly by our partners". Microsoft announced Bing as one of them in 2023, so being indexed by Bing is worth checking, not just Google.
• Each company has its own crawlers. OpenAI uses OAI-SearchBot for search, Anthropic uses Claude-SearchBot, and Perplexity uses PerplexityBot. If your robots.txt or firewall blocks them, those assistants can't read your pages when they search.

What stays the same

Google is blunt about this: "From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO." To appear in AI Overviews, a page has to be indexed and eligible for a normal search snippet, and "there are no additional technical requirements."

So the basics still carry most of the weight:

• Pages that load, can be crawled and are indexed (in Google and Bing).
• One clear page for each important question your buyers ask.
• A good reputation on sites other than your own.
• Writing that answers the question directly instead of circling it.

If your SEO foundations are broken, no GEO tactic will make up for it.

What changes with AI search

1. The result is a sentence, not a blue link

In classic search, success is a position on a page. In AI search, success is being named, in the right context, with the right description. "Quoth is a B2B AI search visibility agency" is a result. So is being listed third in "the best GEO agencies for B2B SaaS are…". Neither shows up in a rank tracker.

2. Other people's pages matter more than yours

When an assistant answers "what's the best X?", it mostly repeats what the pages it read already say. Those are often review sites, comparison articles, forums, videos and encyclopaedias rather than company websites. Studies of AI citations show this clearly, although the mix varies a lot by platform and by month:

• Across ChatGPT, Perplexity and Google AI Overviews, Ahrefs found Wikipedia was the most-cited domain on all three in its June 2025 data.
• In Ahrefs' study of Google AI Overviews across more than 3 million US queries, YouTube and Reddit were the two most-cited domains (June 2026).
• Semrush saw the share of ChatGPT answers citing Reddit fall from close to 60% to around 10% after September 2025, which shows how quickly these patterns can shift.

The practical lesson: find out which pages the assistants cite for your questions, and work on being included there.

3. Consistency is how AI recognises you

AI systems piece together who you are from many sources. If your website, your LinkedIn page, your Crunchbase profile and your directory listings all describe you in the same words, the picture is clear. If they disagree, the AI may blend you with someone else.

We saw this first-hand. On 6 October 2026 we listed Quoth on Crunchbase, using exactly the one-sentence description from our website. Two days later, Google's AI Overview for "quoth agency" described us as "a specialized B2B AI search visibility agency" and cited that Crunchbase profile. One consistent sentence, repeated in the right places, did more than any amount of on-page tweaking would have.

4. Facts and quotes beat keywords

The original GEO paper tested nine ways of rewriting web pages and measured how much more often generative engines used them. Adding statistics and adding quotations performed best, with the best methods improving visibility by up to about 40%. Adding more keywords ("keyword stuffing") gave little to no improvement. It's one study on one benchmark, but the direction makes sense: assistants look for specific, checkable claims they can repeat.

5. You have to measure differently

AI answers change from run to run, and differ by assistant, location and wording. One screenshot proves very little. The useful measure is share of answers: pick a fixed list of buyer questions, ask several assistants the same way every week or month, and count how often you're named, in what position, and which sources were cited. That gives you a trend you can trust.

Popular tactics that don't hold up

• llms.txt files. llms.txt is a proposal for a file that summarises your site for AI agents. Google says Search "doesn't use them", and we found no major AI company confirming that its search uses them. Adding one is harmless, but don't expect it to change your visibility.
• FAQ schema for AI. Google stopped showing FAQ rich results in May 2026 and says there is "no special schema.org markup you need to add" for generative AI search. Write good FAQs for readers; don't add markup expecting AI to reward it.
• Chopping pages into tiny chunks. Google says "there's no requirement to break your content into tiny pieces for AI to better understand it."
• Blocking Google-Extended to stay out of AI Overviews. Google-Extended controls whether Google can use your content for Gemini training and grounding. It doesn't affect Google Search, and AI Overviews are part of Search.
• Mass-produced AI content. Publishing hundreds of thin, AI-written pages to "cover every prompt" risks falling foul of Google's spam policy on scaled content abuse, and it rarely gets cited anyway. Assistants quote pages that say something specific.

How AI is changing SEO and GEO work itself

AI isn't only the channel; it's also changing how the work gets done. Used well, it speeds up the parts that used to be slow:

• Finding the questions. AI tools are good at expanding a handful of buyer questions into the many ways people actually phrase them, and at grouping them by intent.
• Running the checks. Asking five assistants 100 questions every week by hand is tedious. Tools and scripts make it repeatable, which is what makes the trend trustworthy.
• Spotting gaps. Comparing what assistants say about you with what they say about competitors quickly shows missing facts, outdated information and pages that don't exist yet.
• First drafts. AI can produce a structured first draft of an answer page or comparison in minutes.

What AI doesn't do well, and where people still make the difference:

• Original facts. Your own data, prices, customer results and opinions are what make a page worth citing. AI can't make those up, and shouldn't.
• Judgement. Deciding which ten questions matter most for your sales, and which third-party sites are worth approaching, is strategy, not text generation.
• Relationships. Getting included in a respected roundup or review usually takes a person writing to another person.
• Checking. AI-drafted text needs a human to verify every claim before it goes live. An assistant that quotes a wrong number from your site is worse than one that doesn't quote you at all.

What the best GEO agencies do differently

If you're comparing agencies, "best GEO agency" lists won't settle it on their own. Many are written by agencies that rank themselves first. Instead, look for these signs:

1. They measure before they change anything. A baseline on a fixed list of questions, across named assistants, with dates.
2. They work off your website, not just on it. Review sites, comparison articles, publications and communities are where most AI answers come from.
3. They can explain why a competitor is named instead of you, using the sources the assistants actually cite.
4. They don't sell tricks. No guaranteed rankings, no secret prompts, no mass content.
5. They show their own results. A GEO agency should be able to show how visible it is itself, honestly, even if it's new.
6. They report the same way every month, so the trend means something.

For a weighted version you can use on a sales call, see how to choose a GEO agency: a scorecard.

A quick checklist to start this week

1. Check that Google and Bing have indexed your key pages.
2. Make sure robots.txt and your firewall don't block OAI-SearchBot, Claude-SearchBot or PerplexityBot.
3. Write one sentence that describes your company, and use it everywhere: your website, LinkedIn, Crunchbase and directories.
4. List 10 questions your buyers ask before they buy. Ask five assistants and write down who gets named.
5. Note which pages those answers cite. That's your outreach list.
6. Add specific facts to your key pages: prices, numbers, named customers and clear comparisons.
7. Repeat step 4 the same way next month.

Frequently asked questions

Is GEO replacing SEO?

No. GEO builds on SEO. Assistants that search the web still need to find and trust your pages, and Google treats optimising for its AI features as ordinary SEO. What GEO adds is a focus on being named in answers, on third-party sources, and on measuring share of answers instead of rankings.

How long does GEO take to work?

For assistants that search the web, changes can show up within days or weeks once the pages they read change. Our own Crunchbase listing appeared in a Google AI Overview two days after we created it. Changes to what a model knows from training take much longer, and Google AI Overviews generally follow normal search rankings, which take time for new sites.

What is the best GEO agency?

It depends on your company, so be wary of any list that names one winner. The best GEO agency for you is one that measures your current visibility first, explains why competitors are named instead of you, works on the third-party sources AI reads, and reports the same metrics every month. Our scorecard helps you compare agencies, including us, on evidence rather than claims.

Can I do GEO myself with AI tools?

Partly, yes. The checklist above is a good start, and AI tools can help you run checks and draft pages. Where an agency helps is the ongoing work: tracking at scale, getting included on the right third-party sites, and turning the data into a plan. Our guide to getting your SaaS recommended by ChatGPT walks through the do-it-yourself version step by step.