How to get cited by ChatGPT, Claude, and Perplexity

The big shift between SEO and GEO is the role of structure and consensus. Google ranks pages partly by how many other respected pages link to them; AI engines weight how clearly a page answers a specific question and how often the same answer appears across multiple respected sources. The result: pages that are useful, scannable, and referenced by other people's content can get cited by ChatGPT or Perplexity without ever ranking on page 1 of Google for the same query.

The most useful frame for an SEO practitioner in 2026 is: GEO is a second discipline built on top of SEO, not a replacement for it. The pages best-positioned for GEO citations are usually also well-structured for Google, but the reverse isn't always true. A high-domain-authority page that's a wall of paragraphs can rank #1 on Google and never get cited by Claude.

Here are the terms you'll see throughout this guide, in plain English:

GEO (Generative Engine Optimization)

Optimizing content to be discovered, understood, and cited by AI search engines (ChatGPT, Claude, Perplexity, Google AI Mode, Microsoft Copilot). Distinct from SEO because the signals are different: structural clarity, source consensus, and entity recognition outweigh classic ranking signals like backlinks.

AEO (Answer Engine Optimization)

A near-synonym for GEO, sometimes used when the focus is on direct-answer surfaces (Google AI Overviews, voice assistants, featured snippets) rather than LLM chat citations. The tactics overlap heavily.

Citation

When an AI engine names your URL in its answer, typically as a clickable source link. The GEO equivalent of a #1 ranking, except citations can come from pages that don't rank in classic Google search at all.

Crawler / user agent

The bot AI engines use to fetch web content. GPTBot (OpenAI), ClaudeBot (Anthropic), PerplexityBot (Perplexity), Google-Extended (Google's AI training crawler). Blocking these in robots.txt removes you from citation eligibility.

llms.txt

A proposed markdown file at /llms.txt that lists a site's key resources for LLM consumption. Adoption is roughly 10% of domains in 2026. Microsoft and OpenAI crawlers fetch it; Google doesn't.

Brand mention

An unlinked reference to your brand or product elsewhere on the web. Counts as a consensus signal for LLMs even without a backlink, which makes brand-mention strategies a distinct part of GEO from traditional link-building.

Entity

A canonical thing (a product, a company, a concept) recognized by LLMs across sources. Strong entity recognition (clear Wikipedia entry, consistent description across the web, structured data) makes a brand much more likely to be cited.

For the broader SEO picture (where AI helps with classic SEO, where it hurts, what Google's Helpful Content classifier does), see our AI SEO for small business pillar guide. This guide focuses specifically on the LLM-citation half of the discipline.

1. ChatGPT (OpenAI)

   Citation behavior: consensus-driven. Cites when web browsing is active; doesn't cite on training-data answers. Top sources: Wikipedia (~7.8%), Reddit (~12%), and competitor sites that appear consistently across the web. Speed correlation: pages with First Contentful Paint under 0.4 seconds average 6.7 citations versus 2.1 for slower pages. Crawler: GPTBot. Tactical priority: be a consistent presence across multiple respected sources on your topic, not a single perfect page.

2. Claude (Anthropic)

   Citation behavior: doesn't cite by default; will when given source material or when its web search tool fires. Favors structural clarity. Bullet-pointed structured content is 30% more likely to be cited than equivalent paragraph form. Top formats: technical documentation, PDFs, whitepapers, clear definitional content. Crawler: ClaudeBot. Tactical priority: write in clearly structured sections with named concepts, FAQ blocks, and bulleted answer lists.

3. Perplexity

   Citation behavior: real-time retrieval over a proprietary 200+ billion URL index; cites by default. Reddit dominates Perplexity citations at 46.7% of top citations, followed by official documentation. Content can appear in citations within hours or days. Crawler: PerplexityBot. Tactical priority: answer-first paragraphs, structured headers, frequent re-publishing, and active engagement in Reddit threads in your space.

4. Google AI Overviews / AI Mode

   Citation behavior: hybrid. 54% of citations overlap with Google's top 20 organic results; 48% come from outside the top 100 organic. Multi-modal content (text plus images plus video plus structured data) shows the strongest correlation with inclusion (r=0.92 in 2026 research). Trigger rate ~13% of US desktop searches, much higher for informational queries. Crawler: Google-Extended (AI training) plus standard Googlebot. Tactical priority: rank pages on Google first, then layer custom images, video, and schema on top.

1. Clear answer in the first 50 words of each section

   LLMs lift the opening sentence or two of each section as their cited extract. Putting the direct answer in the first 50 words massively raises citation odds across all four engines.

2. Complete schema markup

   Article, FAQPage, HowTo, Organization, BreadcrumbList, DefinedTermSet, Speakable. Schema is rules-based extraction food for LLMs. More schema = more structured signals = more citations.

3. Named author with verifiable credentials

   Real author byline tied to a Person schema entity with sameAs links to LinkedIn, X, or a public profile. LLMs cite content tied to real people more readily than anonymous content.

4. Recent publish or modified date

   Visible on the page and in Article schema. Perplexity and ChatGPT live search both weight freshness. Stale dates suppress citation odds even when content is otherwise current.

5. Two to five outbound links to authoritative third-party sources

   Linking out to primary sources (government, academic, established publishers) signals that the page is part of a verifiable source network, which LLM citation algorithms reward.

The cross-engine traits are not subtle, but they're also not what most SEO advice has emphasized historically. Notice what's NOT on the list: domain authority, backlink count, keyword density, content length. None of those correlate strongly with AI citation. What does correlate: the page being structurally machine-readable, written by an identifiable human, and referenced from other respected sources².

1. Article (with Person author)

   The base layer. Article schema tells AI engines the page is editorial content with a real author. Adding a Person node for the author (with sameAs links to verifiable profiles) strengthens entity recognition.

2. FAQPage

   Q&A pairs in structured form. The single most-cited schema across all four AI engines because FAQ answers map directly to user questions. Use it for every guide that includes a real FAQ section.

3. HowTo

   Numbered step-by-step instructions in structured form. Cited heavily by ChatGPT and Claude for tactical queries. Use for any page describing a multi-step process.

4. DefinedTermSet

   Glossary entries in structured form. Excellent for getting cited on definitional queries ("what is X"). Each term gets a separate citation opportunity.

5. Speakable

   CSS-selector markup telling voice/answer engines which sentences to lift. Underused and high-leverage. Target your direct-answer sentences (TL;DR, section openers, FAQ answers).

6. Organization (with sameAs)

   Your entity profile, linked to Wikipedia, Crunchbase, LinkedIn, or other authoritative entity sources via sameAs. Strengthens brand entity recognition across all AI engines.

A few implementation notes:

• Use multiple schema types per page. A good guide page can carry Article + FAQPage + HowTo + Speakable + Breadcrumb simultaneously. Each one is a different extraction surface for different AI engines.
• Link your Person and Organization entities. Author schema references the Organization via worksFor; Organization links to Wikipedia, Crunchbase, or other authoritative entity sources via sameAs. The entity graph you build internally informs how AI engines model your brand.
• Speakable is underused. Most sites don't implement Speakable. It tells voice and answer engines exactly which sentences to lift, which boosts citation odds for direct-answer content (TL;DR, section openers, FAQ answers). Trivial to add, meaningful upside.

The Reddit play

Reddit is the single most-cited source across ChatGPT, Claude, and Perplexity. The reason is structural: Reddit threads are answer-rich, frequently updated, community- ranked (so the best answers float to the top), and unlikely to be marketing fluff. AI engines treat upvoted Reddit comments as a strong consensus signal.

The Reddit play for GEO is not what most SEO teams do (drop links and run). It's the opposite. Identify 3 to 5 active subreddits in your domain. Contribute thoughtful answers to real questions over weeks and months. Establish a reputation as a knowledgeable voice without spamming your own site. The citation upside compounds because your username starts appearing alongside trusted answers, and your domain becomes referenced in those threads organically over time.

The Wikipedia play

Wikipedia accounts for roughly 7.8 percent of ChatGPT citations and similar shares across the other engines. A Wikipedia presence does two things for GEO: it makes your brand a recognized entity (which LLMs cite more readily), and it provides an authoritative anchor that other sources can reference. If your business is notable enough to qualify for a Wikipedia article (real news coverage, real third-party writeups), create one through Wikipedia's normal process. Don't self-edit; Wikipedia's editors will revert promotional edits and damage your entity signals.

The YouTube play

YouTube transcripts are increasingly used as source material by AI engines, particularly for how-to and explanatory queries. A YouTube channel where you actually talk about your domain provides a structured, attributed, transcribed source that LLMs can cite. The barrier to entry is lower than most B2B teams realize: even rough single-take explainer videos work if the transcript is clear.

The principle behind all three

AI engines scan for consensus before they confidently cite a brand. The more places your brand appears consistently across the web (Reddit threads, Wikipedia, YouTube, industry directories, your own site, partner blogs), the more confidently engines will surface you as an answer. The single most-overlooked GEO investment is the hours spent being present in the source ecosystem outside your own site.

The llms.txt proposal mirrors the structure of robots.txt: a small, plain-text file at a predictable URL that tells AI crawlers what to read and in what order. The companion llms-full.txt provides full text content in machine-readable markdown for ingestion.

The current state of play in 2026:

• Adoption. About 10.13 percent of domains have published an llms.txt file. Up from near-zero a year ago but still niche⁴.
• Google. Has publicly stated it does not use llms.txt for AI features. Google's AI surfaces rely on standard crawling.
• OpenAI and Microsoft. Profound's crawler activity data shows OpenAI and Microsoft bots actively fetching both llms.txt and llms-full.txt files. Models from both companies appear to use the markdown content for live retrieval.
• Other engines. Adoption signals are mixed; the safest assumption is that more engines will start reading llms.txt over time as the standard becomes more established.

The pragmatic recommendation: publish both llms.txt and llms-full.txt. Cost to implement is trivial. Downside is essentially none. Upside is better fidelity in AI engine ingestion when the engines do read it. Treat it as a low-cost infrastructure bet, not a guaranteed visibility lift.

Our own /llms-txt and /llms-full-txt routes auto-generate from the site's guide registry, so adding a guide auto-publishes it for AI consumption. The mechanism is not magic; it's 30 lines of code that maps URLs to descriptions.

The major AI crawlers and the user agents to whitelist:

• GPTBot (OpenAI): crawls the web for ChatGPT training and live retrieval.
• ClaudeBot (Anthropic): crawls for Claude's web search tool and training.
• PerplexityBot (Perplexity): crawls for Perplexity's real-time retrieval index.
• Google-Extended (Google): Google's opt-out signal for AI training. Allowing it permits Google to use your content for AI features.
• cohere-ai, anthropic-ai, ccbot: other AI crawlers worth allowing depending on which engines you want to be visible in.

The minimum-viable robots.txt for GEO is: don't block any of these user agents. If you have a legitimate reason to block one (privacy, copyright, competitive), it's a real trade-off: you opt out of citation in that engine. Most small businesses should allow all major AI crawlers as the default.

How to verify they're actually crawling you

Check your server logs for the user agents above. If you're using Cloudflare, the AI bot management dashboard shows which AI crawlers have requested which URLs. If you're not seeing any AI crawler hits, your site may be too small to be in the crawl queue yet, your robots.txt is blocking them, or your hosting is throwing 503 errors on bot requests. All three are fixable.

Manual tracking (free, 30 minutes a month)

Pick 20 queries that matter to your business. Once a month, run each through ChatGPT (with web search on), Claude (with web tool on), Perplexity (default), and Google AI Mode (default). For each query, note: which pages get cited, which competitors get cited, which Reddit threads or Wikipedia entries get cited. Track the list in a simple spreadsheet over time.

Manual tracking is sufficient for businesses early in GEO. The patterns you spot (which engines cite you, which competitors keep showing up, which content formats win) are the same patterns automated tools surface, just with fewer queries checked.

OtterlyAI ($29/month and up)

Automated tracking of brand mentions and URL citations across ChatGPT, Perplexity, Google AI Mode, Google AI Overviews, Gemini, and Copilot. Includes audit reports and data exports. The lowest-friction starting point when manual tracking is no longer enough. Used by small and midsize teams to monitor their AI search visibility weekly without spending a person's time on it.

Profound (enterprise-tier)

Used by Fortune 500 firms for AI search intelligence with SOC 2 compliance. Its distinctive feature is Prompt Volumes, which show actual AI search demand by topic with demographic breakdowns. Think of it as keyword research for AI search: instead of guessing which queries matter in ChatGPT, you see the real volumes. Overpriced for small businesses; right tool for teams treating GEO as a strategic channel.

What to track over time

• Citation count per engine. How many of your tracked queries cite your pages each month.
• Citation share by competitor. Which competitors keep showing up, and on which queries.
• Source pattern. When you're NOT cited, what sources get cited instead. Reddit threads you should join? Wikipedia entries to update? Industry sites worth contributing to?
• Brand mention frequency. Are you being mentioned in answers without being cited (unlinked mentions)? Those count as consensus signals even without the link.

1. Days 1 to 5: Baseline measurement

   Pick 20 queries that matter to your business. Run each through ChatGPT (with web search), Claude, Perplexity, and Google AI Mode. Note which engines cite your pages, which cite competitors, and which cite Reddit / Wikipedia / industry sites. That's your starting map.

2. Days 6 to 12: Schema and structure upgrade

   Audit your top 10 pages by traffic. Add Article + Person author + FAQPage + HowTo schema where applicable. Restructure section openers to lead with direct answers in the first 50 words. Add 2-5 outbound links to primary sources per page. Add an llms.txt file at /llms.txt.

3. Days 13 to 20: Source ecosystem work

   Identify 3-5 active Reddit threads in your space. Contribute thoughtful answers (not links to your site) that establish you as a knowledgeable voice. Update your Wikipedia presence if relevant. Get listed on 2-3 industry directories that LLMs cite.

4. Days 21 to 27: Publish new content for the gaps

   Write 2-3 new pages targeting queries where you noticed competitors getting cited but you weren't. Apply the cross-engine traits: 50-word direct answer per section, complete schema, real byline, recent date, outbound links. Submit to Search Console for fast indexing.

5. Days 28 to 30: Re-measure and decide

   Re-run the same 20 queries through all four engines. Note new citations, ranking shifts, and any engines you're still invisible on. Decide what worked, what didn't, and which of the next 30-day cycle's queries to prioritize. The compounding starts here.

What this 30-day cycle produces: a real baseline of which queries cite which pages, schema and structural improvements on your top 10 pages, a meaningful presence in 3-5 active source-ecosystem threads, 2-3 new pages targeting competitive citation gaps, and the data to decide what cycle 2 should prioritize. Expect first visible citation lifts in cycle 2 (days 30-60) and compounding through cycle 4 (days 90-120).

If you want the broader picture of how GEO fits inside SEO (Google AI Overviews, E-E-A-T, Helpful Content classifier, the whole stack), our AI SEO for small business pillar guide covers the full landscape.

If you're considering using AI to actually write the content the GEO playbook requires, our Will AI content hurt your SEO? guide covers what works, what gets penalized, and the recovery playbook for sites already hit.

If you'd rather have someone else audit your specific GEO posture and tell you which engines you're invisible on and how to fix it, our free 48-hour assessment sends a written read on your current AI citation footprint, the schema and structural gaps holding you back, and what performance terms we can offer to build and run the engine for you. No sales call.