To rank in ChatGPT, Google AI Mode and Perplexity, a page has to clear three gates in order: the AI crawler must be allowed to fetch it, the search index the assistant draws on must already rank it, and the page must contain a short, specific, quotable answer from a source the model can identify. Most sites that are invisible in AI answers fail the first gate and never find out.

“Generative engine optimisation” (GEO) has become the label for this work, and a cottage industry has grown up around it. Strip away the new vocabulary and most of it is good SEO with three additions: crawler access for AI bots, content written to be quoted, and a clear, consistent identity for your business across the web. This guide covers what the data actually says, what we changed on our own site, and the order to do things in.

How AI search picks its sources

AI assistants do not have a secret index of their own for live questions. When ChatGPT, Perplexity or Google’s AI Mode answers something current, it runs a web search behind the scenes, reads a handful of the results, and writes an answer that cites some of them. That has two practical consequences.

First, ranking in traditional search still matters enormously. Industry studies through 2026 disagree on exact figures, but they agree on the direction: a large share of ChatGPT’s citations overlap with Bing’s top results, and Google’s AI features draw heavily, though not exclusively, on pages Google already ranks. If you are on page four for a query, you are rarely the source an assistant quotes.

Second, ranking is necessary but not sufficient. Of the pages an assistant reads, it cites the ones that answer the question most directly and most specifically. A page that buries its answer under three paragraphs of preamble loses to one that states it in the first sentence.

Assistant Where its live results come from Crawler to allow Sources it tends to favour
ChatGPT search Its own search plus Bing OAI-SearchBot, ChatGPT-User, GPTBot Reference sites, forums, established publishers
Google AI Overviews and AI Mode Google’s own index Googlebot (Google-Extended for Gemini training) YouTube, forums, pages already ranking
Perplexity Its own index plus partner search PerplexityBot, Perplexity-User Recent, specific, well-structured pages
Claude Web search ClaudeBot, Claude-SearchBot, Claude-User Clear primary sources and documentation
Microsoft Copilot Bing Bingbot Whatever ranks in Bing
Source tendencies are summarised from 2026 industry citation studies (Peec AI, Profound and others). Treat them as directional: each engine changes its behaviour often.

Gate one: make sure AI crawlers can actually reach you

This is the cheapest fix and the most commonly missed. When we audited our own site this year, robots.txt allowed every crawler, yet OpenAI’s GPTBot and Meta’s crawler were receiving HTTP 429 “too many requests” errors on every single page. The block was not in WordPress at all. It was a bot-protection setting at the hosting company’s CDN, switched on by default.

Nothing in your analytics will tell you this is happening. You have to test it. Request one of your pages while identifying as each crawler and check the status code:

  • robots.txt: confirm you are not disallowing GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot or Google-Extended, often added by a security plugin or a copied template.
  • CDN and firewall: Cloudflare, Hostinger, SiteGround and others now ship “block AI bots” toggles. Check each one.
  • Rate limits: aggressive limits can block crawlers that request several pages quickly, which is exactly how crawlers behave.
  • JavaScript rendering: most AI crawlers do not run JavaScript. If your content only appears after scripts load, they see an empty page.

Whether to allow AI training crawlers is a business decision. For a service business that wants to be recommended, the answer is almost always yes: being absent from a model’s training data and its search results is a cost, not a protection.

Gate two: rank in both Google and Bing

Because ChatGPT and Copilot lean on Bing, Bing has quietly become worth caring about again. Two hours of work covers most of it:

  • Verify your site in Bing Webmaster Tools (you can import directly from Google Search Console) and submit your sitemap.
  • Turn on IndexNow, which tells Bing and other engines the moment a page is published or updated, instead of waiting to be recrawled.
  • Claim Bing Places alongside your Google Business Profile. Local AI answers draw on both.

Beyond that, the ranking work is the same as it has always been: technical health, pages that genuinely match search intent, and links from relevant sites. Our guide to what a technical SEO audit finds covers the foundations.

Gate three: write pages an assistant can quote

This is where GEO differs most from traditional SEO. A language model building an answer looks for a passage it can lift, compress or cite with confidence. You make that easy with structure, not tricks.

Content pattern Why assistants cite it How to apply it
Answer in the first sentence The quotable passage is easy to find Open with the answer, then explain
Specific numbers Specifics beat generalities when a model compares sources Price ranges, timelines, thresholds, dates
Comparison tables Structured data is simple to extract accurately One table per major comparison
Question-style headings They match how people phrase prompts H2s and H3s written as real questions
FAQ with schema Self-contained question and answer pairs Five to eight genuine questions, marked up
Visible update dates Freshness is weighted for anything time-sensitive Show and maintain “last updated”
Named, accountable source Models favour sources they can identify Clear company details and real authors
Patterns that recur across 2026 citation studies and in our own testing.

Notice what is missing: keyword density, word count targets and “AI-optimised” filler. Pages padded to hit a length rarely get cited, because the useful passage is harder to find. The best-cited pages in most niches are dense, specific and slightly opinionated.

Make your business easy to identify

Assistants answering “who is the best web development company in Kathmandu” are not reading one page. They are assembling a picture of each business from many sources and recommending the ones they can describe confidently. You help by making that picture consistent.

  • One description, everywhere. Write a single sentence that says what you do, where, and for whom, and use it word for word on your site, social profiles and directory listings.
  • Organization schema. Mark up your name, address, contact details, service area and social profiles in JSON-LD so the facts are machine-readable.
  • Third-party mentions. Directory listings, review platforms, industry roundups, forum threads and press coverage are where assistants look for corroboration. A business mentioned only on its own website looks unverified.
  • An llms.txt file. A short plain-text guide at /llms.txt listing your key pages with one-line descriptions. It is an emerging convention rather than a ranking factor, but it costs an hour and gives AI tools a clean map of your site.

A 30-day GEO plan for a small business

Week Focus Tasks
1 Access Test every AI crawler for 200 responses, fix robots.txt and CDN blocks, check content renders without JavaScript
2 Indexing Bing Webmaster Tools, IndexNow, Bing Places, Google Business Profile, sitemap clean-up
3 Identity Organization schema, one consistent description, llms.txt, claim key directory profiles
4 Content Rewrite the opening of your top ten pages to answer first, add tables and FAQs, show update dates
Ongoing after month one: one genuinely useful, specific article a fortnight, and a monthly visibility check.

How to measure AI search visibility

There is no Search Console for ChatGPT yet, so measurement is partly manual. Write down 15 to 20 questions your customers actually ask, such as “how much does a website cost in Nepal” or “web developer in Kathmandu for an Australian company”. Ask each one monthly in ChatGPT, Perplexity, Gemini and Google’s AI Mode, and record whether you are mentioned, whether you are linked, and which competitors appear instead.

Alongside that, watch three signals you can measure properly: referral traffic from chatgpt.com, perplexity.ai and similar domains in GA4; impressions in Bing Webmaster Tools; and AI crawler hits in your server logs. If crawlers are visiting and referrals are growing, the work is landing.

Frequently asked questions

Is GEO different from SEO?

Mostly no. GEO is SEO plus three additions: allowing AI crawlers, writing content that can be quoted directly, and building a consistent identity across the web. A site with weak traditional SEO will rarely do well in AI search, because assistants pull from pages that already rank.

How long does it take to appear in ChatGPT answers?

Fixing crawler access and Bing indexing can show results in weeks for queries where you already rank. Being recommended for competitive queries takes the same months of consistent work as ranking in Google, because it depends on the same authority signals.

Should I block AI crawlers to protect my content?

For a service business that wants customers, blocking usually costs more than it protects. If your content is your product, such as paid research, the calculation is different, and you might allow search crawlers like OAI-SearchBot while blocking training crawlers like GPTBot.

Does llms.txt improve rankings?

Not directly. No major assistant has confirmed using it as a ranking signal. It is a cheap, low-risk way to give AI tools a clean summary of your business and key pages, which is why we publish one, but it is not a substitute for crawler access and good content.

Can I pay to appear in AI answers?

Not in the organic answer itself. Some platforms are testing ads alongside AI answers, but the cited sources are chosen by the system. Anyone promising guaranteed placement in ChatGPT answers is selling something they cannot control.

Do I need a GEO tool?

Not to start. Manual monthly checks across four assistants, plus GA4 referrals and Bing data, are enough for most small businesses. Paid visibility trackers become useful when you are monitoring dozens of queries or competing in a crowded category.

Where to start

Test crawler access today. It takes ten minutes and it is the problem we find most often. If you would rather have someone check the whole chain, from crawler access and indexing to schema and content, tell us your site and we will look at it. Our SEO service and SEO packages cover the ongoing work.

Share this article