How to Rank in AI Search: SEO, AEO, GEO and LLMO Explained

To rank in AI search, get the SEO basics right and then extend them. Your pages must be indexed and open to the right crawlers. They must answer real questions clearly, show first-hand experience and come from a brand the web already recognizes.

AEO, GEO and LLMO are names for that same work on different surfaces. Google now says so in writing: its guide to optimizing for generative AI features calls the work “still SEO” and lists llms.txt files and content “chunking” among things you can ignore.

I have spent more than 12 years in SEO, and I lead the SEO work at Adtric. Below is what I tell clients about AI search, with the numbers my team has seen so far.

SEO, AEO, GEO, LLMO and AI SEO in plain words

None of these terms has an official definition, and most people use them interchangeably. Here is how I use them.

LLMO vs GEO. GEO has a clear origin. It comes from a 2023 research paper, GEO: Generative Engine Optimization, later accepted at KDD 2024. LLMO is looser. People use it for how a model talks about your brand, including answers it gives without searching the web. The practical work is nearly identical.

Is GEO the same as AI SEO or LLMO? Close enough that one plan should cover all of them. Google’s guide names AEO and GEO, then says: “From Google Search’s perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.”

My client work points the same way. Almost all of the core SEO work carries over to AEO. As I wrote in a June 2026 post on the Adtric blog: “They are not separate disciplines. They are SEO, extended.” The extension is real, though. A rank report cannot tell you whether an assistant names your brand, so you have more to measure.

One distinction does matter. Some AI answers come from a live web search, and some come from what a model learned in training. A live search can only cite pages it can fetch, so SEO applies directly. You cannot submit anything to training data. You can only be described widely and accurately on the web, and decide whether training crawlers may read your site.

How Google’s AI Overviews and AI Mode choose sources

Google’s AI features sit on top of its normal search index. The guide says they rely on Google’s “core Search ranking systems to retrieve relevant, up-to-date web pages from our Search index”. Google calls this grounding. I put it bluntly in that Adtric post: “If your site is not retrievable by Google, it is invisible to AI Overviews as well”.

Both features may also use “query fan-out”. Google’s AI features page describes it as “issuing multiple related searches across subtopics and data sources”. One question becomes many searches, and the answer links to pages that support it.

I gave a local example in another Adtric post, Google Said AEO Is Still SEO. A buyer asks for the best ready-to-move 3 BHK under ₹2 crore in Noida Extension. The system may look up projects, prices, possession status and locality reviews separately, then combine them. A page built for one head keyword misses most of those searches.

Google is plain about requirements: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” A page “must be indexed and eligible to be shown in Google Search with a snippet”. In practice, check three things:

• Indexing. If key pages are missing from Google’s index, nothing else on this list matters. My guide to how Google crawls, indexes and decides priority shows how to find and fix them.
• Snippet controls. Google names nosnippet, data-nosnippet, max-snippet and noindex as ways to limit what it shows. Check that none is set by accident on pages you want cited.
• The Search generative AI setting. A Search generative AI control under Settings in Search Console decides whether your content can appear in AI Overviews and AI Mode. Inclusion is the default, and Google says it rolled out to all websites as of August 31, 2026. Check that nobody has switched it to exclude.

How ChatGPT search and Perplexity find your pages

ChatGPT and Perplexity each run a dedicated search crawler. Block it, and your content stays out of that company’s search answers. Their training and user-request bots are separate, and both companies say each setting works independently.

OpenAI’s publisher FAQ says any public website can appear in ChatGPT search. To be included in its summaries and snippets, do not block OAI-SearchBot. OpenAI’s crawler documentation says it can take about 24 hours for a robots.txt change to reach search.

Where ChatGPT gets results matters too. OpenAI’s ChatGPT search help page says it “may share disassociated search queries with third-party search providers such as Bing”. That makes Bing indexing part of AI search work. Verify your site in Bing Webmaster Tools and submit your sitemap there as well.

Perplexity works the same way. Its crawler page says PerplexityBot is “designed to surface and link websites in search results” and recommends allowing it in robots.txt.

Here are the bots worth knowing, as each company describes them:

Robots.txt is not the only gate. Google also asks you to allow crawling in “any CDN or hosting infrastructure”, because a firewall can block a bot that robots.txt allows. OpenAI and Perplexity publish IP ranges for their bots, so you can allow them precisely.

Neither OpenAI nor Perplexity documents how it chooses which pages to cite. Be wary of anyone who claims to know the formula.

What actually helps you get cited

Once the technical doors are open, the work looks like good SEO with less room for shortcuts. These are the five areas I focus on.

Make important pages easy to retrieve

When I audit a site, I act like a visitor first and ask what the site is trying to say. Then I check the technical side, including a site: search on Google to see what is indexed. If Google cannot find and read a page, no AI feature built on its index will use it.

Keep key content in text, not inside images or scripts. Google’s AI features page lists “Making sure that important content is available in textual form” as a best practice.

Answer the questions buyers actually ask

Fan-out means one page rarely wins alone. Buyer questions are getting longer, more local and more personal. In India, many are spoken, and many mix Hindi and English. So my team at Adtric maps every priority service, project or location to a cluster of related questions.

Google says you “don’t need to write in a specific way just for generative AI search”. Microsoft’s Bing Webmaster team says clear headings, tables and FAQ sections help AI systems reference your information accurately. Both point the same way: lead with the answer and write clearly for people.

Show first-hand experience and put a name on it

Google’s guide asks for “non-commodity” content that “provides unique expert or experienced takes that go beyond common knowledge and the ordinary”. It also says: “Don’t just recycle what others on the internet have already said.” In my experience, commodity answers get summarized. Original data, real cases and named experts get cited.

Bylines matter more than they used to. Google’s guidance on helpful content asks: “Do pages carry a byline, where one might be expected?” I said it more bluntly on the Adtric blog: “Pages with named experts win. Pages without lose.”

The strongest asset is data only you have. My bank migration guide on the Adtric blog is built on numbers only my team had. One example is how many of the bank’s most-clicked addresses were PDFs. Nobody can write that page by summarizing other pages.

Make your brand easy to recognize

In my work, AI answers lean toward brands the system already knows. That is why press coverage, reviews and a presence where your category is discussed now feed search visibility. Keep your name, address, phone and description consistent on your site, your Google Business Profile, directories and Wikipedia if you have a page.

Earn mentions from real coverage, real customers and real expertise, not bought ones. If you sell locally or online, keep your Business Profile and Merchant Center details current. Google says both can help your products and services appear in AI responses.

Use structured data where it matches the page

Structured data is not a ticket into AI answers. Google says “there’s no special schema.org markup you need to add”. It still supports rich results, and Google asks that it match the visible text on the page. My team at Adtric adds it to priority pages, and I check it with Google’s Rich Results Test.

What to skip: llms.txt, chunking and bought mentions

Google’s guide has a section called “Mythbusting generative AI search: what you don’t need to do”. It lists five things you can ignore for Google Search:

• llms.txt files and other “special” markup.
• Chunking content into tiny pieces for AI.
• Rewriting content just for AI systems.
• Seeking inauthentic mentions.
• Overfocusing on structured data.

llms.txt is a proposal from September 2024 for a file that helps AI agents use a website. Google says you may keep one for other systems that read it. But it “will neither harm nor help your site’s visibility or rankings in Google Search, as Google Search ignores them.” For other AI search tools, it has little evidence behind it. At Adtric, we tell clients that plainly and skip it.

An earlier version of this article recommended an llms.txt file and 40 to 60 word answer blocks. I have removed both. Google’s current guidance does not support them, and I would rather correct advice than defend it.

If an agency pitches AEO or GEO to you, Google suggests checking it against its guidance on third-party SEO advice. Google doesn’t evaluate third-party services, and its AI guide adds: “No third-party tool has access to our internal ranking or AI systems.”

My own test is simple. SEO is the foundation of AEO, and AEO asks for more than SEO did. An agency that says nothing has changed will under-scope your work. One that sells AEO as a replacement for SEO will oversell it.

What changed in my client work, and what did not

AI search did not make SEO harder. It made it less forgiving. That was the argument of my June 2026 post, and this list comes from it.

Position 1 is the change clients find hardest to accept. On one education client in Delhi NCR, my team saw rankings hold steady for a year while impressions kept climbing. Clicks kept falling, quarter after quarter, starting around the time AI answers began appearing on more of its searches. The SEO had not gotten worse. The answer was simply arriving before the click.

Content volume is the other one. I do not believe you must publish three times a week to rank. One page with facts nobody else has is worth more than a stack of generic posts.

How to measure AI search traffic and visibility

A rank tracker alone cannot show AI visibility. Use these sources together:

My team at Adtric runs that manual check every month. We ask real buyer questions in ChatGPT, Gemini and Google’s AI surfaces, and record whether the brand appears.

What the numbers look like so far

AI assistants still send far fewer visits than Google. But the visitors they send often act. Here is what my team at Adtric has recorded, each with its own window:

• A school with two campuses in Greater Noida West. ChatGPT visits rose from 14 in December 2024 to February 2025 to 168 in June to August 2026. From December 1, 2024 to September 24, 2026, 3.0% of ChatGPT visits included a key event, against 2.2% of Google visits (case study). Of 671 AI-assistant visits to the main campus site from December 2024 to September 2026, ChatGPT sent 563 and Perplexity 69. My guide to SEO for schools covers the rest of that work.
• An Indian manufacturer of transmission towers that sells abroad. ChatGPT sent 86 sessions from March to September 2026, against 45 in the nine months before (case study). We built that site for search from day one, as my SEO checklist for new websites describes.
• A small finance bank that moved to a .bank.in domain. In 2026, the site recorded 720 visits from AI assistants, and 17.4% ended in a submitted Apply Now form. That is 3.4 times the site’s average, as I reported in the migration guide.

These numbers are small, so I treat them as an early signal, not a trend. After the bank’s domain move, my team also asked ChatGPT, Gemini and Perplexity for its official website. In September 2026, all three gave the new domain. We keep asking.

Finally, judge the channel by leads, not clicks. A BFSI client of ours wanted to stop SEO after reading that AI was killing Google. AI Overviews showed on roughly half of his priority searches. But branded search was up year on year, and organic conversions were up. Organic also had the lowest cost per qualified lead of any channel he ran. He kept the retainer.

Where to start this month

Start with access, then answers, then measurement:

1. Check robots.txt and your CDN or firewall rules for Googlebot, OAI-SearchBot and PerplexityBot.
2. Confirm the Search generative AI setting in Search Console is set to include your site.
3. Fix any important page that Google has not indexed.
4. List the questions your buyers ask, and make sure one clear page answers each.
5. Add a named author and first-hand detail to your most important pages.
6. Set up GA4 key events, and watch the AI Assistant and Organic Search channels.
7. Ask ChatGPT, Gemini and Perplexity those buyer questions once a month, and note who they name.

If you want help, my monthly SEO consulting treats Google and AI search as one plan. It includes a monthly check of your buyers’ questions in ChatGPT, Gemini and Google’s AI Mode, noting whether your site is cited. I do not sell AI answers as a separate product. It is the same work, checked in more places.