How to Get Cited by ChatGPT and Perplexity: A 90-Day Plan

Why do some sites get cited and others not?

Sites get cited when an AI tool can fetch them, finds a passage that answers the search it ran, and sees the same facts elsewhere. Sites that miss one of these three conditions are skipped, however good their design or however long they have ranked.

Think of a citation as the end of a short chain. The tool receives a prompt, turns it into one or more searches, reads the pages that come back, and quotes the ones that support its answer.

A page can drop out at each link. Its crawler may be blocked. It may not come back for the rewritten search. Or it may come back but say nothing quotable.

Here is the pattern Digital Magma sees most in audits. A business has a strong homepage and a services page that lists twelve offerings in one paragraph. Neither page answers a single question well, so neither is cited.

The cited competitor often has a plainer site. But it has a page titled with the exact question, an answer in the first lines, a price range, and a few reviews.

Being cited is less about authority in the abstract and more about fit. The page that fits the search wins the line in the answer.

How do ChatGPT and Perplexity find the sources they cite?

ChatGPT and Perplexity both cite live web pages, but they reach them differently. OpenAI says ChatGPT rewrites prompts into queries for search partners. Perplexity documents its own crawler, PerplexityBot, which builds the index its answers link to. Both publish crawler names and IP lists.

Two sources sit behind the table. OpenAI’s help article, Searching the web with ChatGPT, covers query rewriting, citations and location. Perplexity’s Perplexity Crawlers page covers its two agents.

OpenAI gives a concrete example. A question about CCR8 cancer drugs might first become the search “CCR8 immunotherapy drug development 2025”, then a narrower follow-up query.

So your page competes for the rewritten query, which is often shorter and more specific than what the buyer typed. Location can be added too, based on the user’s approximate IP location.

OpenAI’s help page lists Microsoft and Shopify privacy policies under its search providers. That is a hint that classic search indexes still feed the ChatGPT side.

Can both crawlers reach you? Checking access before chasing citations

Citation work starts with access, because a page neither crawler can fetch will never be cited. Check robots.txt for OAI-SearchBot and PerplexityBot groups, then check the firewall, since Perplexity says a WAF may need explicit rules to let its bots through.

The robots.txt part is short. Make sure neither name sits under a “Disallow: /” line, and that no blanket rule for all agents blocks them.

User-agent: OAI-SearchBot
User-agent: PerplexityBot
Allow: /
Disallow: /wp-admin/

If you do not want your content used to train models, block GPTBot separately. OpenAI’s crawler overview says each setting is independent of the others. Our GPTBot entry has the copy-ready rules.

The firewall part catches more sites. Perplexity’s documentation gives Cloudflare and AWS steps: match the user agent and the published IP ranges, then set the rule to allow.

Perplexity also advises refreshing those IP ranges automatically, because the lists change. A rule copied once in 2025 may now block real crawls.

Digital Magma checks citation access in the server or CDN logs as well as in the file. Look for the bot names, then confirm the IPs against each company’s JSON list.

Both companies note that their user-triggered fetchers may skip robots.txt. That matters less for citations, because the search crawlers are what build the index.

What pages get cited? Publishing answers AI can quote

The pages that get cited answer one question each, open with the answer, and carry checkable facts such as dates, prices, steps and named sources. Digital Magma writes these as short, self-contained sections, so any one passage can be lifted into an AI answer and still make sense.

Digital Magma starts every page that should get cited from the buyer’s questions, not your service list. Sales calls, WhatsApp chats and review texts are better sources than a keyword tool.

Each question gets one URL. On that page:

1. Put the answer first. Forty to sixty words under the main heading, written so a stranger understands it out of context.
2. Name the subject. Service, location, audience and year in plain words.
3. Add evidence. A number with its source, a dated example, a step-by-step method.
4. Show who wrote it. A named author with real experience in the field.
5. Keep it current. An updated date, and facts rechecked when the source changes.

Research backs part of this. The GEO paper by Aggarwal and colleagues, accepted at KDD 2024, tested methods such as adding citations, quotations and statistics to source pages.

The authors report that GEO can lift visibility in generative engine responses by up to 40% on their own benchmark. Treat that as a lab result, not a forecast for your site.

Google’s guidance points the same way from a different angle. Its guide to generative AI features asks for “non-commodity” content with a first-hand view, rather than summaries of what others have said.

Our AI citation and answer engine optimization entries go further into the writing pattern.

Does outside proof help you get cited? Reviews, directories, press and forums

Outside proof helps you get cited because AI answers often draw on several pages, and pages about you on other sites can confirm what your own site claims. Reviews, trade directories, local press and real forum answers each add a second voice that agrees with you.

When Digital Magma reviews the sources behind a competitor’s citations, the same types recur:

• Review platforms. Google reviews and industry review sites, with counts and ratings visible.
• Lists and directories. “Best X in Y” lists, chamber and association listings, marketplace profiles.
• Press. Local news, trade magazines, interviews.
• Community threads. Reddit, Quora and niche forums where people recommend a supplier by name.

Quality matters more than volume. Google’s guide warns that seeking inauthentic mentions “isn’t as helpful as it might seem”, because its spam systems filter them.

So skip paid link farms and fake reviews. Ask real customers for reviews, answer real questions under your own name, and pitch real stories to journalists.

Digital Magma’s own example is its Google rating: 5.0 from 100+ reviews. We show it as evidence buyers can check, without claiming it causes any AI citation.

Digital Magma treats consistency as part of getting cited. Keep your facts the same everywhere. A different phone number or business name on a directory gives an AI tool two versions of you to choose between.

What Digital Magma has seen in clients that get cited

The Digital Magma clients that get cited by AI tools are the ones that already earn steady search traffic. Citations followed rankings, and rankings followed pages that answered real buyer questions. We have not seen a shortcut that skipped those steps.

The clearest case is a Canadian welding manufacturer we have worked with for three years. It holds 40+ keywords in Google’s top 10 and is visible in ChatGPT and AI Overviews.

In one three-month period, the site drew about 30K clicks from about 100K impressions. The citations sat on top of that search presence, built from product and buying-guide pages.

A Toronto ENT clinic, a client for four years, shows the local side. It earned about 20K clicks in three months from search, built on condition and treatment pages written for patients.

Neither result came from a “citation package”. Both came from pages that answer what buyers ask, kept accurate over years.

That is the honest limit of what Digital Magma can promise. We can make a site citable and measure the change. The AI tool still decides what to quote.

How do you track citations over time?

Track citations with a fixed prompt list, run on a fixed schedule, with each result logged by date, platform and cited URL. For Microsoft Copilot and Bing, add the AI Performance report in Bing Webmaster Tools, which counts citations directly. ChatGPT and Perplexity offer no such owner report today.

Microsoft introduced AI Performance in Bing Webmaster Tools on 10 February 2026. It shows total citations, average cited pages, grounding queries and page-level citation activity.

Microsoft calls the grounding queries a sample, so read them as direction, not a full list. They are still the closest thing to seeing which searches an AI used to find you.

For ChatGPT and Perplexity, build a manual log. A simple sheet works:

• Prompt, written exactly the same each time
• Platform, date and whether search ran
• Your brand named: yes or no
• Your URL cited: yes or no, and which URL
• Competitors named and cited

Digital Magma runs each Perplexity and ChatGPT prompt more than once. Answers vary between runs, so a single test can mislead in either direction.

Add referral traffic as a third signal. In GA4, sessions from chatgpt.com or perplexity.ai show that cited links are being clicked.

Google’s AI Overviews and AI Mode have their own report now. Search Console’s generative AI performance report shows impressions in those features, and Google says it reached all sites on 31 August 2026.

A 90-day plan to get cited by ChatGPT and Perplexity

A 90-day citation plan gives month one to access and a baseline, month two to answer pages, and month three to outside proof and a retest. Digital Magma uses this split because citations tend to lag the work by weeks, so the plan needs room for that delay.

Days 1 to 30: access and baseline

• Write 15 to 20 buyer prompts. Include price, comparison and “best in city” questions.
• Run each in ChatGPT with search on and in Perplexity. Log the results.
• Open robots.txt and firewall rules to OAI-SearchBot and PerplexityBot. Confirm visits in logs.
• Verify the site in Bing Webmaster Tools and Search Console.

Days 31 to 60: answer pages

• Map each prompt to one URL. Create the missing pages.
• Rewrite the opening of each page so the answer comes first.
• Add dates, figures with sources, and an author box.

Days 61 to 90: proof and retest

• Fix name, address and phone mismatches on profiles and directories.
• Ask recent customers for reviews. Pitch one story to a trade or local publication.
• Rerun the full prompt list and compare it with day one.

Expect uneven results. Some prompts move in weeks, others not at all in 90 days. Keep the list fixed, and judge progress across all prompts together.

Getting cited by AI: frequently asked questions

These are the citation questions marketing managers ask Digital Magma before they hire anyone. The answers rely on what OpenAI, Perplexity, Google and Microsoft publish, and say plainly where those companies are silent.

Related citation questions, sources and author

These pages continue the citation topic from different angles. Digital Magma links live knowledge-base entries directly and marks planned answers until they are published.

• AI citation: what counts as a source link in an AI answer
• Generative engine optimization: the research term behind citation work
• Query fan-out: why one prompt becomes several searches
• Perplexity, PerplexityBot and ChatGPT search (planned entries)
• Why a website is not in ChatGPT, What makes content easy for AI to quote and How to measure AI visibility

Sources: OpenAI’s crawler overview and its help article on web search, Perplexity’s crawler documentation, Google’s guide to generative AI features and its Search Console help pages, Microsoft’s Bing Webmaster blog, and the GEO paper on arXiv. All were checked on 22 October 2026.