---
title: "The ChatGPT and Perplexity Citation Checklist: 18 Checks Before You Rewrite Your Copy"
description: "Competitors getting cited while you stay invisible? Run these 18 checks in bot order: access, readability, answerability, corroboration, then content gaps."
url: "https://groas.com/post/the-chatgpt-and-perplexity-citation-chec"
image: "https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/236f479a-4b6f-4e80-a016-41d74c3e9f24.png"
published: "2026-10-18T05:47:31.919Z"
modified: "2026-10-18T05:47:31.980Z"
---

October 18, 2026 · 9 min read

# The ChatGPT and Perplexity Citation Checklist: 18 Checks Before You Rewrite Your Copy

[Alexander PerelmanHead Of Product @ groas](https://groas.com/author/alexander-perelman)[LinkedIn](https://www.linkedin.com/in/alexander-433793253/)

![Tiny diorama: couriers stopped at a closed boom gate while painters beyond it polish a billboard nobody can reach—rewriting copy behind a locked gate.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/236f479a-4b6f-4e80-a016-41d74c3e9f24.png)

In this article

1. [Stage 1: Can the bots get in?](#stage-1-can-the-bots-get-in)
2. [Stage 2: Can they read an answer on the page?](#stage-2-can-they-read-an-answer-on-the-page)
3. [Stage 3: Do you answer the question they asked?](#stage-3-do-you-answer-the-question-they-asked)
4. [Stage 4: Does anyone else say it?](#stage-4-does-anyone-else-say-it)
5. [Stage 5: Where are you actually absent?](#stage-5-where-are-you-actually-absent)
6. [The one-page version: mark each check pass or fail](#the-one-page-version-mark-each-check-pass-or-fail)
7. [The check people skip, and what it costs](#the-check-people-skip-and-what-it-costs)

On our site, live-answer bots fetched **robots.txt 1,227 times in 30 days**. The homepage got 801 fetches. That is the part most “get cited by AI” advice skips: before a bot can quote your carefully rewritten paragraph, it has to get through the gate. I would run these 18 checks in the order a bot meets the site: **access, readability, answerability, corroboration, then content gaps**. Stop at the first fail. Fixing a paragraph below a locked gate is just paying someone to improve a diary.

## Stage 1: Can the bots get in?

1. **Allow OAI-SearchBot in robots.txt.** Pass if the file does not disallow it on pages you want surfaced. OpenAI [splits the jobs](https://developers.openai.com/api/docs/bots): OAI-SearchBot is used for ChatGPT search, while ChatGPT-User fetches in response to a user request and does not decide search inclusion. Check the rule for the bot doing the job; an allowance for ChatGPT-User is not a substitute for one for OAI-SearchBot.
2. **Confirm ChatGPT-User, PerplexityBot, and Claude-User can fetch the pages that matter.** Pass if live-answer fetches return 200 rather than 403. Perplexity [also splits its bots](https://docs.perplexity.ai/guides/bots): PerplexityBot respects robots.txt, while Perplexity-User generally ignores it and must be controlled at the firewall. If the rules are a copy-paste job, check which bot each rule covers. The [_AI Crawler Swipe File_](https://groas.com/post/the-ai-crawler-swipe-file-copy-paste-rob) has rules for that split.
3. **Check WAF and bot presets before trusting robots.txt.** Pass if Cloudflare Security > Bots, WAF custom rules, and server logs show OAI-SearchBot and PerplexityBot getting through. Those controls can [act before robots.txt is consulted](https://automatelab.tech/blog/al-products/get-website-cited-chatgpt-perplexity-2026-checklist/). A perfect file cannot rescue a request the edge has already dropped. Allowlist verified IP ranges, then confirm the fetch in the log rather than assuming the setting worked.
4. **Render money-page content in raw HTML.** Pass if View Source shows the title, H1, copy, and links without JavaScript. Most AI crawlers read raw HTML without running JavaScript; in one 41-day test across more than 1,000 pages, GPTBot, ClaudeBot, and PerplexityBot [discovered zero pages linked only through JavaScript](https://speedvitals.com/blog/csr-vs-ssr-ai-seo/). Ship SSR or SSG for pages you want cited. A polished page in your browser is not proof that a crawler received its content.
5. Check target URLs in Bing Webmaster Tools. Pass if they are indexed. Bing visibility is worth testing before you rewrite a page for ChatGPT Search, even if the page looks fine in your browser. Treat this as a discoverability check, not a promise that indexing earns a citation. If the URL is absent, deal with that first; better opening copy cannot help a page that is not being found.

## Stage 2: Can they read an answer on the page?

6. Put a direct answer in the first 100 words of each target page. Pass if a buyer can get it in 2–4 sentences without scrolling. Retrieval [lifts short, self-contained spans](https://automatelab.tech/blog/al-products/get-website-cited-chatgpt-perplexity-2026-checklist/), and 44.2% of citations in the cited sample came from the first 30% of page text. Open each H2 with a 40–60-word answer. Do not make the reader, or the extractor, work through an introduction to find the point.
7. **Put key facts in text and tables, not images, tabs, or accordions.** Pass if you can Ctrl+F the price, specification, or step. In the cited comparison, tables were [extracted at 81% versus 23% for the same content in prose](https://automatelab.tech/blog/al-products/get-website-cited-chatgpt-perplexity-2026-checklist/); FAQ sections were cited at 2.3 times the base rate. Use plain HTML for facts you need a bot to lift. A hero image can look excellent and still be a terrible place to hide a price.
8. Give pricing, locations, and specifications crawlable homes. Pass if each important detail appears on a URL a bot can reach and in a passage that makes sense on its own. Perplexity [cites specific passages](https://upgrowth.in/how-to-get-cited-by-perplexity-ai/), not the impression your whole site leaves. Lead sections with a complete answer, keep paragraphs to 2–4 sentences with one verifiable idea, and use question-shaped H2s where they fit. “As mentioned above” is not an answer when the passage travels alone.
9. **Match organization and product schema to visible copy.** Pass if validator output agrees with what a visitor can see: name, price, availability, location, and ratings. Check the sitemap too; it should list the URL and be referenced in robots.txt. Do not make structured data say one thing while the page says another. If you spend $20k a month but bury the service price in a hero image and the steps in a tab, the bot still has no clean passage to quote.

## Stage 3: Do you answer the question they asked?

10. Give each buyer question one obvious URL. Pass if a prompt you want to appear for has a page that answers it completely, rather than scattering the answer across three service pages. Take a question about who handles emergency calls within 60 minutes in your city. If each part lives somewhere else, a competitor with one complete page is easier to cite. This is a page-planning check, not an instruction to publish a new page for every keyword variation.
11. Publish an honest comparison or _best-for_ page. Pass if at least one page names alternatives, explains who you are best for, and says who should choose someone else. Answer bots get asked _best X for Y_. A page that admits a tradeoff gives them something usable; a page that claims you are best at everything does not. Name the decision a buyer is making instead of filling another page with category claims.
12. Add dated, specific figures that a model can quote. Pass if relevant pages give a date and a figure, such as a starting price, a response time in minutes, or a measured CPA change over a stated period. “AI-powered growth” gives an extractor nothing to lift. Precision does not mean making up a number: use one the business can stand behind. I used to write the vague version. It never got cited.

![A robot hand stamping PASS on a website gate while ignoring the paragraphs behind it](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/52bb0b5d-35e1-447a-839f-c8a51f8b96d5.png)

## Stage 4: Does anyone else say it?

13. Match your identity across directories and review sites. Pass if your homepage and the relevant Google Business Profile, Yelp, Clutch, or G2 listings agree on your name, category, core claims, and where you operate. Check the wording, not just the logo. A business described differently on every third-party page gives a bot conflicting material to work with. This is not a call to join every directory; it is a check on the listings you already use.
14. Earn genuine mentions where Perplexity looks for your prompts. Pass if you can name three non-competitor pages it cites in your niche that mention you accurately. Across the cited 680-million-citation sample, Reddit was Perplexity’s [most-cited source at 6.6%](https://upgrowth.in/how-to-get-cited-by-perplexity-ai/), followed by YouTube at 2.0%. A third-party page may be cited instead of yours. First list the pages Perplexity actually uses for your questions; then work on being represented accurately there.
15. **Track ChatGPT and Perplexity separately.** Pass if your log has a column for each engine rather than one blended “AI visibility” score. In a study of 1,429 answers to 88 contractor questions, LinkedIn appeared [246 times in Perplexity versus six in ChatGPT](https://www.baadigi.com/blog/ai-search-engines-cite-different-sources-study); YouTube appeared 214 times in AI Overviews versus four in ChatGPT. Domain overlap was only 11–12%. A citation in one engine does not establish a win in the other.

![Clay cross-examination with business website claims opposite review sites](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/cac2129e-8faa-42e3-be43-3d3204a8e8d0.png)

## Stage 5: Where are you actually absent?

16. Build a fixed list of real buyer prompts. Pass if you have 20–30 questions drawn from sales calls, support tickets, and search terms, rather than a keyword export renamed “AI strategy.” The [_60 Buyer Prompts Swipe File_](https://groas.com/post/the-buyer-prompt-swipe-file-60-questions) is a starting point if you do not have your own log. A keyword list tells you what you rank for. A prompt list lets you test where you get named.
17. Run each prompt repeatedly in ChatGPT and Perplexity. Pass if every prompt has at least three runs per engine, with the cited domains recorded. A [repeatable mention audit](https://groas.com/post/run-your-own-ai-mention-audit-a-repeatab) can include standard and Pro Search modes. Log the winner, the passage pulled, and whether you appeared at all. One answer is a sample, not a trend; if you cannot repeat the prompt and compare the output, you cannot tell whether a change helped.
18. Call something a gap only when competitors repeat and you never appear. Pass if you can point to fixed prompts where competitors are cited repeatedly and you get zero citations. That is the [gap test practitioners use](https://upgrowth.in/how-to-get-cited-by-perplexity-ai/): run the same list monthly, rather than treating one omission as a verdict. Five misses in a row may point to a missing page, a missing proof point, or a missing third-party mention. Diagnose which before commissioning more copy.

![Infographic checklist showing the five citation failure stages in order, from access to gaps](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/1ccb04f5-d3bc-430a-ac6a-1071c901009a.png)

## The one-page version: mark each check pass or fail

Copy this into a doc and run it top to bottom. **Pass means you tested it today**, not that someone configured it last year. Stop at a failure; the checks below it may not yet tell you anything useful.

1. OAI-SearchBot allowed in robots.txt.
2. ChatGPT-User, PerplexityBot, and Claude-User fetch live pages with a 200 response.
3. Cloudflare Bots, WAF rules, and logs confirm verified bots get through.
4. Money pages show title, H1, copy, and links in View Source.
5. Target URLs appear in Bing Webmaster Tools.
6. Direct answer appears in the first 100 words and under relevant H2s.
7. Key facts appear in text or tables, not only images, tabs, or accordions.
8. Pricing, locations, and specifications appear on crawlable URLs.
9. Organization and product schema matches visible copy; the URL is in the sitemap.
10. Each buyer question has one clear URL that answers it.
11. An honest comparison or _best-for_ page names alternatives.
12. Relevant figures are specific, dated, and supportable.
13. Name, category, location, and core claims match across relevant listings.
14. Three non-competitor pages Perplexity cites in your niche mention you accurately.
15. ChatGPT and Perplexity results have separate log columns.
16. The audit uses a fixed list of 20–30 real buyer prompts.
17. Each prompt has three runs per engine, with citations and passages logged.
18. A gap means repeated competitor citations and zero appearances for you.

If several fail, fix them in bot order rather than difficulty order: firewall and robots.txt, indexing and raw HTML, then answer spans. Once all 18 pass, use the repeated prompt results to choose the missing asset: a page that answers the exact question in one quotable span, or a third-party mention that backs the claim. Build that asset and rerun the same prompts. Do not mistake a single fresh answer for proof that the work landed.

![Hourglass pouring sand onto a blocked website gate while fresh articles wait behind it](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/8d176f45-bb23-422b-adc5-a4a4eae3d3e3.png)

## The check people skip, and what it costs

3. **Confirm the WAF lets verified bots through.** This is the one people skip because robots.txt looks correct. The edge can still drop OAI-SearchBot or PerplexityBot before either reads it. Then the team pays to rewrite answers no bot can fetch. Check the logs for a 200 before buying another round of copy work. If you want prompts, citations, and crawler visits [tracked with actions logged](https://groas.com/mcp) rather than handed back as another monthly FAQ recommendation, that is the work groas runs under. But the order stays the same: clear the gate first, or the rewrite costs you time without giving a bot a page to read.

## Frequently Asked Questions

### Does allowing ChatGPT-User in robots.txt get my site into ChatGPT search results?

No. OpenAI splits the jobs: OAI-SearchBot is used for ChatGPT search, while ChatGPT-User fetches pages in response to a user request and does not decide search inclusion. You need an allowance for OAI-SearchBot itself in robots.txt on the pages you want surfaced.

### Does Perplexity-User follow robots.txt rules?

Perplexity splits its bots: PerplexityBot respects robots.txt, while Perplexity-User generally ignores it and has to be controlled at the firewall instead. So if you want to block Perplexity-User, robots.txt is not the lever; use your firewall or WAF settings.

### My robots.txt allows AI bots, so why are they still blocked?

WAF and bot presets such as Cloudflare Security > Bots or custom WAF rules can act before robots.txt is consulted, so a request can be dropped at the edge even when the file is correct. Allowlist the verified IP ranges and then confirm in your server logs that OAI-SearchBot and PerplexityBot actually get a 200 response rather than assuming the setting worked.

### Can ChatGPT and Perplexity crawlers read JavaScript-rendered content?

Most AI crawlers read raw HTML without running JavaScript. In one 41-day test across more than 1,000 pages, GPTBot, ClaudeBot, and PerplexityBot discovered zero pages linked only through JavaScript. Ship SSR or SSG for pages you want cited, and confirm in View Source that the title, H1, copy, and links appear without JavaScript.

### Is it better to put pricing and specs in tables or in paragraph text for AI citations?

Tables. In the comparison cited, tables were extracted at 81% versus 23% for the same content in prose, and FAQ sections were cited at 2.3 times the base rate. Put key facts in plain HTML text or tables; an image or accordion can look good and still leave the bot nothing to lift.

### Where on the page should the direct answer go for AI citations?

Put a direct answer in the first 100 words of each target page, and open each H2 with a 40-60-word answer a buyer can get in 2-4 sentences without scrolling. Retrieval lifts short, self-contained spans, and in the cited sample 44.2% of citations came from the first 30% of page text.

### Do ChatGPT and Perplexity cite the same websites?

No. In a study of 1,429 answers to 88 contractor questions, LinkedIn appeared 246 times in Perplexity versus six in ChatGPT, and YouTube appeared 214 times in AI Overviews versus four in ChatGPT. Domain overlap was only 11-12%, so track the two engines separately in your logs rather than using one blended AI visibility score.

### When does a missing AI citation actually count as a content gap?

Only when competitors are cited repeatedly and you never appear. Build a fixed list of 20-30 real buyer prompts from sales calls and support tickets, run each prompt at least three times per engine in ChatGPT and Perplexity with citations logged, and call a gap only where you get zero citations across repeated runs. Five misses in a row may point to a missing page, proof point, or third-party mention; diagnose which before commissioning more copy.

## Pay For Results, Not For Hours

Businesses buy the outcome, agencies resell it, and groas answers for it either way.

[See If You Qualify](https://groas.typeform.com/to/xC1bQNUT)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/the-chatgpt-and-perplexity-citation-chec#webpage","url":"https://groas.com/post/the-chatgpt-and-perplexity-citation-chec","name":"The ChatGPT and Perplexity Citation Checklist: 18 Checks Before You Rewrite Your Copy","description":"Competitors getting cited while you stay invisible? Run these 18 checks in bot order: access, readability, answerability, corroboration, then content gaps.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/236f479a-4b6f-4e80-a016-41d74c3e9f24.png"},"breadcrumb":{"@id":"https://groas.com/post/the-chatgpt-and-perplexity-citation-chec#breadcrumb"},"dateModified":"2026-10-18T05:47:31.980Z","mainEntity":{"@id":"https://groas.com/post/the-chatgpt-and-perplexity-citation-chec#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/the-chatgpt-and-perplexity-citation-chec#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"The ChatGPT and Perplexity Citation Checklist: 18 Checks Before You Rewrite Your Copy","item":"https://groas.com/post/the-chatgpt-and-perplexity-citation-chec"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/the-chatgpt-and-perplexity-citation-chec#article","url":"https://groas.com/post/the-chatgpt-and-perplexity-citation-chec","mainEntityOfPage":{"@id":"https://groas.com/post/the-chatgpt-and-perplexity-citation-chec#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"The ChatGPT and Perplexity Citation Checklist: 18 Checks Before You Rewrite Your Copy","description":"Competitors getting cited while you stay invisible? Run these 18 checks in bot order: access, readability, answerability, corroboration, then content gaps.","datePublished":"2026-10-18T05:47:31.919Z","dateModified":"2026-10-18T05:47:31.980Z","image":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/236f479a-4b6f-4e80-a016-41d74c3e9f24.png"},"author":{"@id":"https://groas.com/author/alexander-perelman#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"keywords":"pass, page, perplexity, chatgpt, copy, cited, bot, checks","wordCount":1841,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/alexander-perelman#person","name":"Alexander Perelman","jobTitle":"Head Of Product @ groas","description":"Ex Goldman Sachs and Ex Stanford Computer Science","image":"https://groas.com/media/blog/4f17cac3a81acc224cc1b2cfab8f94bbc295a3089d13e8421125771ef8d4064f.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"},"sameAs":["https://www.linkedin.com/in/alexander-433793253/"]},{"@type":"FAQPage","@id":"https://groas.com/post/the-chatgpt-and-perplexity-citation-chec#faq","mainEntity":[{"@type":"Question","name":"Does allowing ChatGPT-User in robots.txt get my site into ChatGPT search results?","acceptedAnswer":{"@type":"Answer","text":"No. OpenAI splits the jobs: OAI-SearchBot is used for ChatGPT search, while ChatGPT-User fetches pages in response to a user request and does not decide search inclusion. You need an allowance for OAI-SearchBot itself in robots.txt on the pages you want surfaced."}},{"@type":"Question","name":"Does Perplexity-User follow robots.txt rules?","acceptedAnswer":{"@type":"Answer","text":"Perplexity splits its bots: PerplexityBot respects robots.txt, while Perplexity-User generally ignores it and has to be controlled at the firewall instead. So if you want to block Perplexity-User, robots.txt is not the lever; use your firewall or WAF settings."}},{"@type":"Question","name":"My robots.txt allows AI bots, so why are they still blocked?","acceptedAnswer":{"@type":"Answer","text":"WAF and bot presets such as Cloudflare Security > Bots or custom WAF rules can act before robots.txt is consulted, so a request can be dropped at the edge even when the file is correct. Allowlist the verified IP ranges and then confirm in your server logs that OAI-SearchBot and PerplexityBot actually get a 200 response rather than assuming the setting worked."}},{"@type":"Question","name":"Can ChatGPT and Perplexity crawlers read JavaScript-rendered content?","acceptedAnswer":{"@type":"Answer","text":"Most AI crawlers read raw HTML without running JavaScript. In one 41-day test across more than 1,000 pages, GPTBot, ClaudeBot, and PerplexityBot discovered zero pages linked only through JavaScript. Ship SSR or SSG for pages you want cited, and confirm in View Source that the title, H1, copy, and links appear without JavaScript."}},{"@type":"Question","name":"Is it better to put pricing and specs in tables or in paragraph text for AI citations?","acceptedAnswer":{"@type":"Answer","text":"Tables. In the comparison cited, tables were extracted at 81% versus 23% for the same content in prose, and FAQ sections were cited at 2.3 times the base rate. Put key facts in plain HTML text or tables; an image or accordion can look good and still leave the bot nothing to lift."}},{"@type":"Question","name":"Where on the page should the direct answer go for AI citations?","acceptedAnswer":{"@type":"Answer","text":"Put a direct answer in the first 100 words of each target page, and open each H2 with a 40-60-word answer a buyer can get in 2-4 sentences without scrolling. Retrieval lifts short, self-contained spans, and in the cited sample 44.2% of citations came from the first 30% of page text."}},{"@type":"Question","name":"Do ChatGPT and Perplexity cite the same websites?","acceptedAnswer":{"@type":"Answer","text":"No. In a study of 1,429 answers to 88 contractor questions, LinkedIn appeared 246 times in Perplexity versus six in ChatGPT, and YouTube appeared 214 times in AI Overviews versus four in ChatGPT. Domain overlap was only 11-12%, so track the two engines separately in your logs rather than using one blended AI visibility score."}},{"@type":"Question","name":"When does a missing AI citation actually count as a content gap?","acceptedAnswer":{"@type":"Answer","text":"Only when competitors are cited repeatedly and you never appear. Build a fixed list of 20-30 real buyer prompts from sales calls and support tickets, run each prompt at least three times per engine in ChatGPT and Perplexity with citations logged, and call a gap only where you get zero citations across repeated runs. Five misses in a row may point to a missing page, proof point, or third-party mention; diagnose which before commissioning more copy."}}]}]}
```