Citation & Source Influence

How to Get Cited by ChatGPT (2026): A Data-Backed Guide

Q: Does it matter whether I allow GPTBot or only OAI-SearchBot?

Yes, they do different jobs. OAI-SearchBot is the crawler that matters for ChatGPT search: OpenAI says sites that block it won't be shown in search answers. GPTBot governs whether your content may be used to train OpenAI's foundation models. Many publishers allow OAI-SearchBot and block GPTBot, and that split is legitimate. ChatGPT-User, which fetches pages on a user's request, is a third agent, and OpenAI says robots.txt rules may not apply to it.

To get cited by ChatGPT in 2026, your content usually needs to clear seven eligibility bars at once: it is shaped for passage-level retrieval (lists, tables, definitions, sourced data), carries verifiable authority signals, is reachable by OAI-SearchBot and indexable by the search providers ChatGPT uses, shows clear recency, is corroborated by other independent sources, has source-layer presence on the surfaces ChatGPT currently cites (since August 2026, official vendor pages, documentation, and help centers far more than community forums), and respects engine-specific differences (ChatGPT is not Perplexity). No single tactic gets you cited, nothing guarantees it, and ChatGPT's paid ads are a separate placement that OpenAI says does not influence answers.

Published 2026-05-22 · Updated 2026-09-26

Questions this guide answers

How do I get my brand cited by ChatGPT?
What content does ChatGPT cite?
How does ChatGPT choose sources?
Which crawler does ChatGPT search use?
Do ChatGPT ads affect which sources get cited?
What types of content get cited in AI answers?

Direct answer

The seven patterns below are the stack. The sections before them explain the evidence behind it and what changed on ChatGPT in August 2026.

The data: what answer engines cite in a B2B category

Before tactics, the evidence. Our 36,268-citation study measured every source cited by 5 engines (ChatGPT, Perplexity, Google AI Overviews, Gemini, and Claude) on 44 buyer-comparison prompts in the AEO category from May 12 to June 11, 2026. The shares below combine all 5 engines and predate August 2026, when ChatGPT changed what it cites (next section). They describe the shape of AI citation in a B2B category; they are not a ChatGPT-only ranking.

Source type	Share of all citations
Other (long-tail blogs, niche and practitioner sites, GitHub, untracked vendors)	45.1%
Vendor sites (tracked AEO, SEO, and marketing-software vendors)	26.1%
UGC (Reddit, YouTube, forums)	11.2%
Editorial (trade press, news, reference, including Wikipedia)	10.3%
Not in the source-type breakdown	7.3%

What this says, honestly

Four things stand out, and the playbook below leans on all of them.

First, owned content alone won't carry you. Vendor sites drew about a quarter of citations combined, but no single vendor's own site earned more than 2.4%. The most-cited domains were Reddit, Wikipedia, and TechRadar. If your AEO strategy is "publish more on our own blog," the data says that is a partial strategy at best, though, as the next section shows, ChatGPT now leans more on official pages than it did.

Second, the source layer matters as much as the publishing layer. Reddit, Wikipedia, and YouTube alone were about 14% of all citations in the window. None of them are owned media; they are community and reference surfaces where you earn presence rather than control it, and the community half is now engine-dependent.

Third, the long tail is the largest bucket. 45.1% of citations went to domains outside the most-cited list. Engines don't pick the same five domains for every prompt; they pull from a wide distribution, which changes how you target source-layer work.

Fourth, the pattern holds outside our category. A June 2026 academic study of 167,551 brand-grounded citations across 128 brands and three models (Perplexity Sonar Pro, Gemini 3.1 Pro, and GPT-5.4) found 85.7% of citations pointed to sites the brand did not own, with Wikipedia the most-cited domain in 11 of 12 languages and YouTube second in its Northern and Baltic dataset. The exact mix depends on the prompts and the definitions; the direction does not.

What ChatGPT stopped citing in August 2026

In the week of Aug. 10, 2026, ChatGPT all but stopped citing Reddit. Tracker data reported by Search Engine Journal showed Reddit's share of ChatGPT search citations falling from 3.83% (July 18 to Aug. 7) to 0.52% (Aug. 14 to 17). In the monitoring workspace behind our citation study, Reddit went from 23.4% of ChatGPT's citations in mid-July to 0.1% from Aug. 14 to Sept. 17, while Google AI Overviews held steady and Perplexity and Gemini, where Reddit was already a small share, stayed small (under 2%). Our baseline is higher than the tracker figure, most likely because the prompts are B2B buyer comparisons; the direction is the same.

Three things follow for anyone optimizing specifically for ChatGPT.

Naming a source is not the same as reading it. Citation data shows what ChatGPT names, not everything it read, so it cannot tell you whether community discussion still shapes answers. The licensing that lets an engine retrieve a source is separate from the product choice to show it underneath the answer. A community thread describing your product wrongly could still influence an answer while being invisible in every citation report you run, so a clean citation dashboard is not proof that community discussion isn't hurting you.

The share moved toward pages you can build. The leading explanation in the trade press is a retrieval change: from Aug. 8, ChatGPT sharply increased fan-out queries that use the "site:" operator, which routes it toward specific, often official, domains (Search Engine Journal has the caveats, including that the steepest drop came six days later). That fits what we measured: in our category workspace, vendors' own sites took a much larger share of ChatGPT's citations after the change, while user-generated sources' share collapsed. If ChatGPT scopes part of its retrieval to specific domains, the question is whether your own documentation answers the query when it arrives.

ChatGPT also shows fewer citations per answer. In the same measurement, citations shown per answer fell from about 33 to about 9, roughly three quarters. A brand whose ChatGPT citation count dropped in August did not necessarily lose ground; the denominator shrank for everyone. Track share, not raw counts, across that date.

What to do differently, specifically for ChatGPT

The seven eligibility patterns below did not change. What changed is where the marginal hour goes.

Make your documentation and help center answer buyer questions, not just user questions. These page types gained ChatGPT citation share. Most vendors write docs for existing customers and leave comparison, pricing-logic, and use-case questions to marketing pages that read as sales copy.
Assume a "site:yourdomain.com" query. If ChatGPT scopes a query to your domain, it needs a page that answers that question directly, with a matching title and a canonical URL. Thin, redirecting, or JavaScript-only pages fail this quietly.
Keep community work, but re-underwrite it. Google AI Overviews still cites Reddit at about its old rate in our data, and buyers read community threads directly. Budget community work as influence and as citations on the engines that still attribute it, not as a ChatGPT citation you can promise.
Measure per engine. An aggregate citation number across engines would have hidden this entirely: one engine lost nearly all of its Reddit citations while the others did not.

What ChatGPT search actually does, and what it doesn't

A lot of confusion comes from conflating two ChatGPT modes that behave very differently.

ChatGPT without search answers from the model's training data. There is no live retrieval, no citation event, and no URL list. You can't "get cited" by the base model in a given week; what matters there is whether your brand and positioning are represented in training data at all, a slower, model-release-cadence problem.

ChatGPT search is the answer engine you want to target. It runs when ChatGPT decides to search the web for an answer, or when a user asks it to. OpenAI does not publish its full ranking system; its documentation and observed behavior point to a pipeline with an eligibility bar at each layer:

Search layer. OpenAI says ChatGPT search sometimes partners with other search providers, rewriting a question into one or more targeted queries that it sends them; the provider privacy policies its help center links are Microsoft's and Shopify's. The Information reported in August 2025 that OpenAI also used Google results obtained through the scraping firm SerpApi for some timely queries. The practical reading: be indexable in both Bing and Google, and don't rely on either alone.
Crawler layer. OpenAI runs three user agents that do different jobs. OAI-SearchBot surfaces sites in ChatGPT search results; sites that block it won't be shown in search answers, though they can still appear as navigational links. GPTBot crawls content that may be used to train OpenAI's foundation models. ChatGPT-User visits a page when a user's request calls for it, and because those visits are user-initiated, OpenAI says robots.txt rules may not apply. If you allow both OAI-SearchBot and GPTBot, OpenAI may use one crawl for both purposes, but they are controlled separately in robots.txt, and OpenAI says a robots.txt change takes about 24 hours to reach search.
Retrieval layer. For each query, ChatGPT search retrieves candidate passages, not whole pages, and ranks them. This is closer to passage-level retrieval-augmented generation than to a ten-blue-links ranking, and since August 2026 part of it is scoped to specific domains with "site:" queries.
Synthesis and citation layer. The model writes the answer from the retrieved passages and attaches source links: the citations you care about. OpenAI says it ranks results on multiple factors and that placement is not guaranteed.

Where ChatGPT ads fit

ChatGPT has carried ads since Feb. 9, 2026, when OpenAI began showing clearly labeled "Sponsored" units below answers for US users on the Free and Go tiers, and OpenAI is now rolling ads out to those tiers in select other regions; Plus, Pro, Business, Enterprise, and Edu accounts don't see them (OpenAI). OpenAI says ads do not influence the answers ChatGPT gives, and the business reached a $1 billion annualized run rate by Aug. 31, 2026.

For citation strategy, that means two separate channels: an ad buys a labeled placement next to the answer, not a citation inside it. Nothing in this guide can be bought through the ads product, and nothing in the ads product changes the patterns below.

The 7 citation-eligibility patterns

These are patterns we see associated with citation events in our own data and in published research. Treat them as correlations, not levers with a known effect size: a July 2026 survey of 45 GEO studies found no reviewed technique with a stable, longitudinal, cross-platform causal effect on discoverability.

1. Content shape AI can extract

ChatGPT search retrieves at the passage level, not the page level, so pages built from clean, self-contained blocks give the retriever more to lift. The test: if one block were pasted into an answer on its own, would it still make sense? Narrative-only essays that only work top to bottom give ChatGPT fewer extractable passages. Two limits apply. This is specific to ChatGPT-style retrieval: Google says its AI features need no special chunking (Google's guide). And shape is not a substitute for relevance: a 2026 study of 252,000 citation trials found formatting-only edits had little impact, while topical relevance drove citation.

Shapes that lift cleanly:

Definition paragraphs: a 2–4 sentence block that opens with the term and answers "what is X?" in plain language.
Sourced lists with parallel structure.
Tables with one row per item and consistent columns.
Visible FAQ blocks with real questions and answers, not FAQ schema added to pages without a visible FAQ (see the anti-patterns).
Numerical claims with units and dates, such as "36,268 citations across 44 prompts, May 12 to June 11, 2026."

2. Authority signals AI can verify

ChatGPT search and its peers look for signals that a claim is verifiable, not just that it sounds confident. Authority in this sense isn't backlink Domain Rating; it's a traceable path from claim to evidence. Pages with a strong evidence trail can be cited on modest domains, and pages on strong domains with no evidence trail can be passed over. Four signals show up repeatedly:

A named author with a real bio and, ideally, a credentials trail.
Inline citations to primary sources: government data, academic papers, and original company research.
Visible publish and updated dates.
Primary research or proprietary data that the page introduces, such as our citation study.

3. Crawler accessibility

If OAI-SearchBot can't fetch a page, the page is not in ChatGPT's candidate pool. The minimum setup: allow OAI-SearchBot and the search crawlers of the other engines you care about (PerplexityBot, Claude-SearchBot and Claude-User, and Googlebot, which Google's AI features use), keep important pages server-rendered or pre-rendered, and don't put content you want cited behind a login. GPTBot is a separate, independent decision about training data: allow it if you're comfortable with training use, block it otherwise.

Common own-goals we still see in 2026:

Blocking OAI-SearchBot in robots.txt, often alongside GPTBot, for reasons no one on the current team remembers.
Auth-gated content that returns a login wall to crawlers; much of a company's best content can sit behind one.
Heavy client-side JavaScript that renders the visible answer only after hydration.
Firewall or CDN rules that block OpenAI's published search-bot IP addresses; OpenAI's help center lists allowing that traffic as a condition of eligibility.

4. Recency

For time-sensitive prompts ("what are the best X tools in 2026?"), ChatGPT search tends to favor fresh sources. OpenAI doesn't publish its refresh timing, so make freshness visible and real:

A visible publish date in the article header, not only in structured data.
A visible "last updated" date on evergreen pages that have genuinely been refreshed.
A year in the title for buyer-guide content, refreshed when the content changes.
A real refresh cadence: change the date only when the substance changes, since a new date on old content is easy to detect and erodes trust.

5. Multi-source corroboration

In our observation, a claim that appears across several independent sources gets cited more readily than one that exists on a single site. The uncomfortable implication for content marketers: a claim often isn't cited until it has been picked up elsewhere. A novel statistic in your blog post is less likely to be cited than the same statistic after a trade publication and a reference site have repeated it.

The tactical move is to earn the corroboration: publish original research, pitch it to trade press, contribute in the communities your buyers use, and update Wikipedia only where notability and its conflict-of-interest rules allow. Corroboration can matter even where a source isn't named, because citation reports show what an engine names, not everything it weighed. We cover the mechanics in earn LLM citations through content-source match.

6. Source-layer presence

The source layer is the set of surfaces ChatGPT trusts that you don't own: Wikipedia, YouTube, G2, industry forums, Reddit, and, increasingly on ChatGPT since August 2026, other companies' official pages and documentation.

In our citation study, Reddit, Wikipedia, and YouTube alone were about 14% of citations before August, and the effect is larger than the share, because source-layer mentions also feed the corroboration in pattern 5. When an engine finds you discussed by real users, referenced in a tutorial, and listed in reference sources, the evidence that you are a real option accumulates. The visible attribution on ChatGPT changed in August; the value of being discussed accurately did not.

This is why "publish more on our own domain" plateaus fast: in our study no single brand's own site passed 2.4% of category citations. Earned and community sources are most of the iceberg. Our Reddit and G2 community sources playbook covers how to build that presence without spamming.

7. Engine-specific patterns

Not every shape or source works equally on every engine. ChatGPT, Perplexity, Google's AI surfaces, Claude, and Gemini retrieve differently, and August 2026 showed that one engine can change what it names within a week while the others don't. The cheat sheet below is directional; measure your own category per engine before acting on it. We cover ChatGPT's specifics in more depth in optimize for ChatGPT Search.

Engine-by-engine citation differences (cheat sheet)

Directional, as of September 2026, and not absolute. The Reddit and YouTube rows trace to measured data (our citation study and published research); the other rows are practitioner judgment, not measurements. The Reddit row is dated deliberately: it moved on one engine in about a week in August 2026, which is the clearest example of why a cheat sheet like this needs a measurement behind it rather than a rule of thumb. An "AI citation strategy" that doesn't break down by engine is leaving accuracy on the table.

Pattern	ChatGPT	Perplexity	Google AIO	Claude	Gemini
Wikipedia weight	Mixed since Aug 2026	High	Med	Med	High
Reddit weight (as of Sept 2026)	Near zero since Aug 2026	Low in our B2B data	Med; held through Aug 2026	None	Low in our B2B data
YouTube weight	Low	Med	High	None in our data	High
Editorial (trade press, reference)	High	Med	Med	High	Med
G2 / Capterra (B2B SaaS)	Med	High	Low	Med	Low
Strong Google organic rank required	Loose	Loose	Tight	Loose	Med
Recency sensitivity	High	Very high	Med	Med	Med

The owned + earned + community recommendation

Given how small any single brand's share of category citations is (2.4% at most in our study), no brand wins AI citations by publishing on its own blog alone. You need three layers, weighted differently than most teams weight them today. The split below is our starting recommendation, not a measured optimum.

Layer 1: Owned (about 35% of effort). Reference-quality pages on your own domain, structured for passage retrieval, with verifiable authority signals, and in particular documentation and help-center pages that answer buyer questions. This layer gained weight after August 2026, because those are the page types that gained ChatGPT citation share.

Layer 2: Earned editorial (about 35% of effort). Coverage in trade publications (TechRadar and vertical trade press), inclusion in third-party comparisons and buyer guides, and citations in industry research. This is what shows up as Editorial, and much of Other, in the citation distribution.

Layer 3: Community (about 30% of effort, down from about 40% before August 2026). Wikipedia where notability allows, Reddit and industry forums (real participation, not promotion), YouTube tutorials, and G2 or Capterra reviews and answers. The reweighting is deliberate: since ChatGPT stopped citing community sources in August 2026, part of this layer's return arrives as influence and as citations on other engines rather than as ChatGPT attribution, and the hours it gives up belong in Layer 1.

The deeper treatment of how to run this 3-layer model is in owned, earned, and community sources for AI.

Anti-patterns: what NOT to do

A regular cycle of tactics sounds plausible and doesn't hold up. The common thread: engines reward verifiable, authentic, well-structured content, and shortcuts that fake those signals get ignored, filtered, or penalized. Rule these out of your roadmap and you'll save quarters of wasted effort.

llms.txt files as a citation lever. Google says its search does not use llms.txt, and John Mueller said in January 2026 that Google's own llms.txt files are not an endorsement. Publishing one is harmless housekeeping, not a way to earn citations.
AI-only schema or "GEO schema." There is no special markup that AI engines reward; Google says so explicitly. Use standard Schema.org markup (Article, Product, Organization, and FAQ only where a visible FAQ exists) and stop there.
FAQ schema on pages without visible FAQs. Google's policy is that structured data must match visible page content. Adding FAQ markup to a page with no visible FAQ is a policy violation, not a shortcut.
One page per fan-out query, or mass programmatic pages. Generating thousands of near-identical pages, one per long-tail variant, is what Google calls scaled content abuse. Near-duplicate pages also compete with each other for the same passage.
Buying brand mentions or paying for "AI seeding." Inauthentic source-layer presence ages badly and creates platform-policy risk; Google's spam policies now cover attempts to manipulate generative AI responses. Earned mentions compound; bought ones decay.
"Guaranteed AI citation lift" pitches. No vendor can guarantee a citation on a specific query, because no vendor controls the retriever. Anyone offering a guarantee either misunderstands the system or is selling mention manipulation. Decline both.

Run a free audit on your own brand

The fastest way to find out what ChatGPT actually cites in your category, and which of these seven patterns you're failing, is to measure your own brand against the questions your buyers ask.

Start Free (free, no credit card, email verification): the free workspace runs 10 buyer prompts on ChatGPT and returns your mention rate, your share of voice against competitors, and the sources ChatGPT cited, plus 1 content audit of a page. Paid plans track more engines daily and rerun the same prompts to verify whether a change you shipped moved the answer; see pricing.

Updated 2026-09-26. Citation data: the SolCrys citation study of the AEO category (36,268 citations, 44 prompts, 5 engines, 2026-05-12 to 2026-06-11) and its August 2026 follow-up; external sources are linked inline and listed below.

Sources

FAQ

How long does it take to see ChatGPT citations after publishing?

OpenAI says a robots.txt change takes about 24 hours to reach ChatGPT search, but it doesn't publish crawl or refresh timing beyond that. In our experience, citations on competitive queries take weeks to months, because corroboration (pattern 5) takes time to build. If you track a fresh prompt set, give it a full quarter of repeated runs before judging.

Do I need to be on Wikipedia to get cited by ChatGPT?

Not strictly, but it helps. Wikipedia was the second most-cited domain in our citation study (1,489 citations, behind Reddit) and the most-cited domain in 11 of 12 languages in a June 2026 academic study. A Wikipedia presence also corroborates what you publish elsewhere. The caveat: Wikipedia has real notability standards and conflict-of-interest rules. You can't create a page for a brand that doesn't meet them, and trying is counterproductive.

Does it matter whether I allow GPTBot or only OAI-SearchBot?

Yes, they do different jobs. OAI-SearchBot is the crawler that matters for ChatGPT search: OpenAI says sites that block it won't be shown in search answers. GPTBot governs whether your content may be used to train OpenAI's foundation models. Many publishers allow OAI-SearchBot and block GPTBot, and that split is legitimate. ChatGPT-User, which fetches pages on a user's request, is a third agent, and OpenAI says robots.txt rules may not apply to it.

What about Perplexity citations specifically?

Perplexity attaches explicit citations to most claims, weighs primary sources and fresh pages heavily, and often cites more sources per answer than ChatGPT does after August 2026 (about 15 versus about 9 in our measurement). Reddit was a small share of Perplexity's citations on our B2B buyer prompts both before and after August (1.5% and 0.2%), so don't assume the consumer-category reputation of "Perplexity loves Reddit" holds for B2B. Optimizing for Perplexity overlaps with ChatGPT on the basics (crawlability, structure, authority) and differs in which sources it favors, so measure it separately.

Will Google AI Overviews cite my page if it doesn't rank in Google organic?

Less likely than on ChatGPT or Perplexity. Google says AI Overviews and AI Mode use its Search index and core ranking systems, and that optimizing for them is still SEO, so pages that already rank are the usual candidates. ChatGPT search draws on third-party search providers and its own crawler, so its relationship to Google rankings is looser. Plan separate tracks for each engine.

Can I pay to get cited by ChatGPT?

No. ChatGPT ads, which began in the US on Feb. 9, 2026 and are rolling out to select regions on the Free and Go tiers, appear as labeled "Sponsored" units below the answer, and OpenAI says ads do not influence the answers ChatGPT gives. An ad buys a placement next to the answer, not a citation inside it. Citations come from the eligibility patterns in this guide.

Should I worry about the ChatGPT base model (no search) at all?

Yes, on a longer time horizon. What the base model "knows" about your brand depends on training data, which changes on a model-release cadence (months, not days). The faster lever is ChatGPT search citations; the slower one is base-model representation. Work both, and expect different timelines.

Does Reddit still help me get cited by ChatGPT?

Not as a visible citation. ChatGPT all but stopped citing Reddit in August 2026 and had not resumed as of mid-September. Citation data can't show whether community discussion still shapes what ChatGPT says, so treat that as unknown rather than assumed. Reddit still earns visible citations on Google AI Overviews, and buyers read it directly. The honest framing: community work buys influence with buyers and attribution on some engines, not a ChatGPT citation.

Free ChatGPT visibility check

See where AI answers skip your brand — then fix it, free

Start a free workspace with your domain: 10 buyer-intent prompts through ChatGPT show where you are mentioned, cited, or skipped, and who gets recommended instead. A free content audit in the same workspace hands you the first fix to ship.

Start Free

Free · No credit card · About 5 minutes

Related guides

Technical Readiness

Search, Agent, or Training? How to Set AI Crawler Access in 2026 Without Blocking Your Own Citations

Which AI crawlers to allow, what Cloudflare's Sept. 15, 2026 rules changed, and how to block AI training without losing your AI citations.

Citation & Source Influence

How AI Answer Engines Choose Sources: The 7 Signals We've Mapped

AI engines like ChatGPT, Perplexity, Google AI Overviews, and Claude choose sources using overlapping but distinct signals. This guide maps the 7 signals that drive citation eligibility and the engine-specific weighting differences.

Citation & Source Influence

Owned, Earned, and Community Sources in AI Answers: A 3-Layer Strategy

AI answers draw on owned, earned, and community sources, and since August 2026 the mix differs sharply by engine. How to balance the 3 layers, engine by engine.

Citation & Source Influence

How to Earn LLM Citations Without Becoming Spam: The Content-Source Match Framework

AI engines preferentially cite 5 specific content patterns. This guide breaks down the content-source match framework, the 12 examples of cited content, and the slow-burn strategy that compounds without manipulation.

Citation & Source Influence

Reddit, G2, and Forums: How to Win the Community Source Layer for AI Citations

AI engines read Reddit, G2 and niche forums heavily - but which engines cite them changed in August 2026. The practitioner playbook for earning community citations without becoming spam, with the 7 rules of native engagement and a per-engine scorecard.

AI Engine Optimization

How to Optimize for ChatGPT Search: The 2026 Practitioner Guide

ChatGPT Search uses Bing's index, OpenAI's crawlers, and on-demand fetches. This guide breaks down the five ranking signals, the crawler access checklist, and the content patterns that get cited in ChatGPT answers.

Citation & Source Influence

Citation Gap Audit

A 5-step framework to identify which sources AI engines cite for your competitors but not for your brand, and the recovery actions for each gap type.

Citation & Source Influence

Schema Markup for AI Search: What Structured Data Actually Does (and Where It Stops)

Schema markup helps AI engines parse and classify your brand, and makes you eligible for richer treatment. It does not, on its own, get you cited or trusted. The honest, complete guide to structured data for AEO: what it does, what it doesn't, which types matter, and how to ship it so it moves the answer.