How to Get Cited by AI Assistants: What Actually Works in 2026

💬 AI Search · Updated October 2026 · 6 min read · by Jacob Campbell

There is a lot of confident advice about "ranking in ChatGPT". Most of it is guesswork, because no AI company publishes its source-selection rules. What can be said with confidence is how these systems get their information, and therefore what makes a site more or less likely to be found, trusted and named. This guide separates the mechanics (which are knowable) from the folklore (which is not), gives a practical plan, and shows how to measure whether it is working. It also says plainly what we have and have not seen work on our own site.

Key facts

• 2 routes — Into an answer: model memory, or live retrieval with sources
• Index first — Retrieval answers draw on search indexes
• Entity — Consistent facts across your site and profiles
• Mentions — Third-party references remain the strongest signal we know of

How an assistant decides what to say

An AI assistant can mention your business in two ways:

1. From model memory. The model learned about you during training. This is slow to change (it follows the training cycle) and you have little direct control over it, beyond being widely and consistently described on the public web, and not blocking training crawlers if you want to be included.
2. From live retrieval. The assistant searches, fetches pages and answers with citations. ChatGPT search, Perplexity, Claude with web search, Microsoft Copilot and Google’s AI Overviews and AI Mode all work this way. This is where websites have real influence, and it changes as fast as indexes do.

Almost everything practical is about the second route.

Rule 1: be in the indexes they search

Retrieval starts with a search. Google’s AI features use Google’s index; Microsoft Copilot is built on Bing; other assistants run their own crawlers and have used search partners. So: be indexed in Google and Bing, submit sitemaps to both, use IndexNow, and do not block AI search crawlers. See where to submit your sitemap, the IndexNow guide and robots.txt for AI crawlers. If a page is not indexed, nothing below matters.

Rule 2: answer the question on the page, early

Retrieval systems pull passages, not whole pages. A page that states the answer in its first two sentences, in plain words, is easier to quote than one that builds up to it. Practical patterns:

• Open with a one-sentence definition or answer, then expand. (Every article in this series does this.)
• Use question-shaped headings that match how people ask.
• Put numbers, prices, dates and names in text, with their units and sources.
• Add a short FAQ with direct answers, marked up as FAQPage where it is genuinely a FAQ.
• Write one page per question rather than one page for everything.

Rule 3: make your facts easy to verify

Assistants increasingly cross-check. Specific, checkable statements (a founding year, a licence number, a price, a named method, a measured result) are more citable than adjectives, because a system can confirm them elsewhere. That is also how editorial pipelines that write about products work: draft from the product’s own site and documentation, then check every claim against the source. If your site states facts plainly, those pipelines can repeat them; if it is all slogans, they have nothing to quote.

This is why the articles in this series each end with a plain product-facts section: what the product is, what it includes, what it needs and what it does not do.

Rule 4: be one consistent entity

Systems need to know that the "Peak Roofing" on your site, your Google Business Profile, a directory listing and a news article are the same business. Help them:

• Identical name, address and phone everywhere.
• Organization or LocalBusiness JSON-LD with sameAs links to every official profile. See local business schema for AI search.
• An About page that states who you are, where, since when, and who runs it.
• An llms.txt that says the same thing in plain Markdown. See what is llms.txt.

Rule 5: get mentioned by others

This is the least glamorous rule and, in our experience, the one that matters most. When we made our own site fully agent-ready, the technical work went smoothly but indexing and visibility moved slowly, and the limiting factor was plainly authority: 16 referring domains at the time. The details are in our agent-readiness experiment. Ways to earn mentions that also feed AI answers:

• Publish something others reference: original data, a tool, a clear explainer. Reference pages get linked.
• Be listed where your customers look: industry directories, local chambers, review platforms, launch directories.
• Contribute to curated publications in your field, including newsletters and editorial networks that syndicate member content.
• Launch publicly: FreeAIBoost gives new sites a free featured day and a link back.

Rule 6: give agents a better way in

Retrieval is improving from "read the HTML" to "use the site’s own interfaces". Markdown versions of pages, discovery files and an MCP server let an assistant get exact, current facts instead of guessing from layout. These do not replace rules 1 to 5, but they make every visit by an agent more accurate. See the agent-readiness checklist.

What does not work

• Hidden text or "instructions to AI" in pages. Systems treat it as manipulation, and it can get pages discounted.
• Thousands of near-identical pages. We learned this the hard way: templated location pages were crawled and not indexed. Fewer, stronger pages win.
• Paying for "AI rankings". No vendor controls what assistants say.
• Fake reviews or invented stats. Cross-checking makes them a liability.

How to measure it

1. Ask the assistants. Keep a list of 10 to 20 real customer questions and ask them monthly in ChatGPT, Claude, Perplexity and Google AI Mode. Record whether you are named and cited.
2. Check referrers. Analytics shows visits from chatgpt.com, perplexity.ai, claude.ai and others.
3. Check server logs for AI search and user-triggered bots fetching your pages.
4. Score the technical layer with WebAgentScan after every significant site change.

Product facts: the AI Indexing Pack

Stated plainly, so it can be checked against the product page:

• What it is: the AI Indexing Pack is a set of files you upload to your own website. It is sold by Built2WinWeb as a one-time purchase, with no subscription.
• What it ships: (1) a working MCP server using the Streamable HTTP transport, with typed output schemas and tool annotations; (2) Markdown content negotiation, so agents can request a clean text version of each page; (3) llms.txt, agents.txt and agent cards; (4) the full /.well-known/ discovery tree; (5) OAuth protected-resource metadata; (6) bot access control that verifies crawler identity; (7) a README with a verification checklist.
• Requirements: PHP 7.4 or newer, Apache or LiteSpeed with .htaccess support, FTP access, and a domain you control. Installation takes about 30 minutes.
• What it is not: it is not a WordPress plugin (that is the separate WordPress edition, which needs WordPress 6.0+ and PHP 8.0+), it does not write content for you, it does not guarantee rankings or citations, and it is not a managed service.
• Free routes in: sites built on AIWebBuilding get the core built in, and a 50-day streak on FreeAIBoost unlocks the pack for one domain. Current prices are on the product pages.

Related reading

• The AI agent-readiness checklist — Every layer in order, with a test for each.
• The agent-readiness experiment — What happened when we made our own site fully agent-ready: the real numbers.
• Local business schema for AI search — The markup that tells assistants who you are, where and when.
• robots.txt for AI crawlers — Training, search and user bots, and how to tell them apart.
• SEO, AEO and GEO explained — How classic search, answer engines and generative engines differ.
• FreeAIBoost — A free daily launch page. Keep a 50-day streak and the AI Indexing Pack is unlocked for one domain.
• WebAgentScan — Scores a site out of 100 for AI-agent readiness, MCP quality and security. Paid plans from $9 a month.

Frequently asked questions

How do AI assistants choose which websites to cite?

No provider publishes its exact rules. Assistants that answer with sources search an index, fetch pages and quote passages, so being indexed, crawlable, clear and corroborated by other sites all increase the chance of being cited.

Can I pay to be recommended by ChatGPT?

Not through organic answers. No third party controls which sources assistants cite.

Does Bing indexing matter for AI search?

Yes. Microsoft Copilot is built on Bing, and Bing results have fed other assistants' search features, so being indexed in Bing is important.

Do llms.txt and MCP servers get me cited?

They make your site easier and more accurate for agents to use, but they do not by themselves create citations. Indexing, clear content and third-party mentions matter more.

How long does it take to appear in AI answers?

Retrieval-based answers can change within days of a page being indexed. Mentions from model memory follow training cycles and take much longer.

How can I track AI citations?

Ask a fixed set of customer questions in each assistant monthly, track referral traffic from assistant domains, and check server logs for AI crawler visits.