How to Get Your Brand Cited by ChatGPT and Perplexity

Getting cited by an AI engine is two separate tests: being retrieved, and then being chosen. Most brands pass the first and fail the second without ever knowing it. In an AirOps analysis of 548,534 pages retrieved across 15,000 ChatGPT prompts, reported by Search Engine Land in March 2026, only 15% of retrieved pages were cited. The other 85% were fetched, read, and discarded before a word of the answer was written.

Here is what decides which group you land in, and the playbook we use.

How ChatGPT and Perplexity actually pick sources

Neither engine is reading your site in real time out of goodwill. Both follow roughly the same pipeline: take the question, expand it into several sub-queries, fetch candidate pages from a search index, and write an answer grounded in the passages that best match.

Three consequences follow.

Which index you are in matters more than most people expect. ChatGPT’s search draws heavily on Bing. A Search Engine Land case study in April 2026 followed a well-reviewed hotel that surfaced in just 1.5% of 68 runs of the same prompt, and found Bing rankings aligned with ChatGPT mentions far more closely than Google rankings did. If you are not in Bing Webmaster Tools, start there.

The question that retrieves you is rarely the one the user typed. Engines fan a prompt out into many sub-queries. In the AirOps data, 15,000 prompts became 43,233 queries, 32.9% of cited pages appeared only in those expanded results, and 95% of the expanded queries had no traditional search volume.

Two different crawlers, two different jobs. OAI-SearchBot retrieves for ChatGPT search and ChatGPT-User fetches a page when someone asks about it directly, while GPTBot collects training data. PerplexityBot indexes for Perplexity. Blocking the training crawler does not remove you from search answers, and allowing the training crawler does not get you cited. The distinction is explained in LLM SEO.

What the citation research actually shows

Three findings are worth designing around.

Answer capsules win. Search Engine Land’s November 2025 analysis of the content traits large language models quote most found that short, self-contained answer capsules were the strongest avenue to citation, and that 34.3% of cited posts combined both an answer capsule and original or owned data. The same analysis flagged heavy link density inside those capsules as a drag on citation.

Mentions beat links. In 2026 correlation analyses, branded mentions tracked AI Overviews visibility at roughly 0.664, against about 0.218 for backlinks. Third-party review profiles showed roughly a threefold citation lift.

Stuffing actively hurts. The peer-reviewed GEO paper at ACM SIGKDD 2024 found keyword stuffing reduced generative visibility by around 9% on its benchmark metric.

Put together: write a clean, quotable answer, put a fact of your own inside it, do not bury it in links, and make sure other people say the same thing about you.

The playbook

1. Confirm you can be fetched

Check three layers, in this order: robots.txt rules for each retrieval crawler, your CDN or firewall bot settings, and your server logs for successful fetches. Cloudflare began blocking known AI crawlers by default for new domains in July 2025, so a permissive robots.txt proves nothing on its own. Details in robots.txt for AI crawlers.

2. Get indexed in Bing, not just Google

Verify the site in Bing Webmaster Tools, submit your sitemap, and use IndexNow so new URLs are pushed rather than waited for. This is the single most under-done technical step for ChatGPT visibility.

3. Render your content on the server

Many AI crawlers do not run JavaScript. View source, not the browser inspector: if your main text is not in the initial HTML, those crawlers see an empty page. See JavaScript rendering and AI crawlers.

4. Write an answer capsule under every heading

Lead each section with a direct, self-contained answer of one to three sentences, then explain. Name the subject rather than writing “it”, because the passage will be read without the paragraphs around it. This is the highest-leverage writing change you can make, and it is covered in content chunking for AI.

5. Put a verifiable fact inside the capsule

A figure, a date, a named source, or your own data. Claims that can be checked are safer for a model to repeat, and original data is what the citation research found paired with capsules in the highest-performing pages.

6. Keep links out of the capsule itself

Put your internal links in the supporting paragraphs rather than inside the answer sentence. Link density inside answer capsules was flagged as a drag in the same analysis, and it makes a passage messier to lift.

7. Make your entity unmistakable

A model that is unsure what you are will cite someone it is sure about. State your category, location, founding year, and people in plain sentences on one canonical page, then mark it up with one Organization node carrying a stable identifier and a short, genuine sameAs list. See the entity SEO guide and Organization schema for AI search.

8. Cover the question space, not the keyword

Since a third of citations arrive through expanded queries with no search volume, breadth of genuine coverage beats precision targeting. Build the cluster: a pillar plus pages answering each real follow-up question. See topic clusters that AI engines cite.

9. Earn independent corroboration

Complete review profiles, accurate directory listings, and inclusion in the category roundups that rank in both Google and Bing. This is slow and it is the part most brands skip, which is exactly why it differentiates. See digital PR for AI search.

10. Measure presence, not just traffic

Most AI answers are zero-click, so traffic undercounts the result. Run a fixed prompt set on a monthly cadence and log presence, citation, accuracy, and which competitors appear. Methods in measuring brand visibility in ChatGPT and share of model.

ChatGPT and Perplexity are not identical

Note one asymmetry worth knowing: PerplexityBot respects robots.txt, while Perplexity-User, which fetches when a user asks about a specific page, generally does not, because it acts on a live request rather than crawling. More in how Perplexity chooses sources.

What does not work

• Publishing more thin pages. Volume without independence is not corroboration.
• Keyword stuffing. Measurably negative, per the research above.
• Hidden text or instructions aimed at crawlers. Against Google’s spam policies, and treated as manipulation.
• Manufactured community posts. Detectable, against platform rules, and damaging to the brand it is meant to help.

The citation checklist

• Retrieval crawlers are allowed in robots.txt and at the CDN.
• The site is verified and indexed in Bing, not only Google.
• Main content is in the initial HTML without JavaScript.
• Every section opens with a self-contained answer.
• Each answer carries a sourced, dated fact.
• Links sit outside the answer sentence.
• One Organization entity with a stable identifier and genuine sameAs list.
• The cluster covers the real follow-up questions.
• Review profiles and directory listings are complete and consistent.
• A monthly prompt set logs presence, citation, and accuracy.

Where citation work fits

Being cited is not a trick to apply to one page. It is the visible result of being retrievable, being quotable, and being corroborated, in that order. The first is technical and fast, the second is editorial and fast, the third is slow and compounds.

If you want to see which of the three is failing for your brand today, request an AI visibility audit: we run your buyer questions across ChatGPT, Perplexity, Gemini, and Google AI Mode, log who gets cited instead of you, and hand back the prioritized gap list. The ongoing work is our GEO service.