Skip to content
The Super Panel Pricing

Blog

ChatGPT SEO: how to get cited by ChatGPT

The short version

ChatGPT SEO is mostly not SEO: you cannot make ChatGPT cite you, and Ethan Smith of Graphite reports only about 35% of ChatGPT citations overlap with Google’s results. What you can do is remove the reasons it cannot: be crawlable, answer directly near the top of a page, be specific enough to quote, and be consistent about who you are.

How to get cited by ChatGPT

Getting cited by ChatGPT is not a setting you enable. Models select sources for reasons nobody outside the labs fully observes, and any guide claiming a reliable method is describing a correlation at best.

What can be said honestly is which conditions make citation possible, which formats get quoted, and how to measure whether your visibility is moving. That is what this page covers.

Can you make ChatGPT cite your website?

Short answer

No. Nobody controls model output, and any provider guaranteeing citations is selling something they cannot deliver. What you can control is whether your page is crawlable, parseable, specific, and formatted so an answer can be lifted from it cleanly.

This distinction matters commercially. It is the difference between buying work on inputs you can verify and buying a promise about someone else’s system.

Start with crawlability, because nothing else works without it

  • Confirm the page is indexed and reachable without JavaScript execution
  • Check robots.txt does not block the crawlers you want to permit
  • Keep the answer in HTML text rather than inside an image or a script-rendered widget
  • Make sure the page loads fast enough that a fetch does not time out

A page that a search crawler struggles with is a page an AI retrieval system will struggle with too. Conventional technical SEO is the prerequisite, not the alternative.

The citation loop

How a page becomes quotable, step by step

Step 1 Fix theprompt set Step 2 Answer in the first 55 words Step 3 Make the entities explicit Step 4 Structure it for machines too Step 5 Re-run and diff
Step 1: Fix the prompt set. Write down the questions you want to be the answer to, and run them against the assistants on a fixed schedule. Without a stable prompt set there is nothing to measure against.
The loop as we run it as a service. Nobody can promise a citation; what can be committed to is the monitoring and the restructuring.

What format gets quoted?

Short answer

Short, self-contained answers placed immediately under a question-shaped heading. Thirty to sixty words that stand alone without the surrounding paragraphs are far easier to lift than an argument that builds across a page.

This is why answer capsules exist on well-optimized pages, including the ones on this site. The format is not decoration: it makes the useful sentence extractable.

Ethan Smith, CEO of Graphite and an SEO practitioner since 2007, has discussed publicly how LLM citation favors content that answers directly rather than content that performs expertise. The formatting implication is straightforward.

In the same interview he reports measuring citation overlap across thousands of questions: roughly 35% between ChatGPT and Google organic results, and roughly 70% between Perplexity and Google. If that holds, most of the answer surface is not covered by ranking work alone.

Ethan Smith is cited as a public source and is not affiliated with this business. Figures as stated by him on Lenny’s Podcast, 14 September 2025: https://www.youtube.com/watch?v=iT7kq-R3Gjc

You can get mentioned by a citation tomorrow and start showing up immediately... So early stage companies can win.
Ethan Smith, CEO of Graphite, 18 years in SEOLenny's Podcast, Sep 2025Cited source. Not affiliated with The Super Panel.
Why this matters
The genuine opening here: unlike organic rankings, citation does not appear to require years of accumulated domain age.

Be specific enough to be worth citing

  • Use numbers, dates and named entities rather than general claims
  • State a position instead of surveying every possible view
  • Cite your own sources, because verifiable pages are safer to quote
  • Answer the question that was asked, not the adjacent one you would rather answer

Generic content is not penalised so much as ignored. If your sentence could have come from any of two hundred pages, there is no reason for a model to pick yours.

What we monitor

Prompt coverage at each AI search tier

Starter 20 prompts Growth 60 prompts Scale 150 prompts
Prompt coverage behaves like a keyword set: it decides how much of the question space you can actually see. It does not decide whether a model cites you.
Published scope from the AI search service page, tier by tier, so the jump between them is visible. Scope is what is committed to; citation by any assistant is not something anyone can commit to.

Does schema markup help with AI citation?

Short answer

Schema does not force a citation, but it removes ambiguity about what a page is, who published it and what it describes. Clear entity data makes a page easier to interpret, and easier interpretation is a precondition for being selected.

Organization, WebPage, FAQPage and Article markup cover almost everything a normal site needs. Marking up content that is not visible on the page is a violation, not a shortcut.

Two printed sheets on a desk with the same short paragraph bracketed on each

How do you measure AI citations?

Short answer

Pick a fixed set of real questions buyers ask, run them monthly across ChatGPT and Perplexity, and record whether you were cited, mentioned or absent, plus which sources appeared instead. A fixed prompt set is what makes month-over-month comparison possible.

Exclude your own brand name from the tracked set. Being cited when someone searched for you proves very little.

Changing the prompts mid-program destroys the comparison, which is the main thing being bought.

The Reddit strategy is the obvious strategy, which is just to be an actual user of Reddit. Make an account, say who you are, say where you work, and give a useful answer.
Ethan Smith, CEO of Graphite, 18 years in SEOLenny's Podcast, Sep 2025Cited source. Not affiliated with The Super Panel.
Why this matters
Unglamorous and slow, and still the most reliable route into the sources assistants quote from.

What ChatGPT SEO actually means

Short answer

ChatGPT SEO is the only phrase in this whole category with measurable search volume. Generative engine optimization, LLM visibility, AI search visibility and get cited by AI all return no data at all.

That is worth pausing on. An industry has produced an entire vocabulary faster than anyone has started searching for it.

It also tells you how much of the surrounding language is supply-side. The work underneath is narrower than the terminology suggests, and the rest of this page is what it actually consists of.

SEO for ChatGPT and the other answer engines comes down to three things: can the crawler fetch the page, does the answer format let a passage be lifted cleanly, and is there anything in it worth citing as a source.

Where this sits against ordinary SEO

  • Crawlable and renderable, for both, with nothing blocked to GPTBot or any other user agent
  • Answer-first structure: lead with the direct answer rather than a wind-up, and skip the filler
  • Specificity, which is what actually gets lifted, rather than restating what everyone says
  • Structured data for entity clarity, which helps parsing and not persuading

The overlap is substantial. A page that is crawlable, well structured and specific is easier for both a search engine and a model to use.

So this belongs alongside ordinary SEO rather than instead of it. Buying ChatGPT SEO while the fundamentals are broken is the wrong order, and the writing service that produces genuinely specific pages does more for citation odds than any markup does.

If you want to be the best credit card, you pay Forbes and then you are the best credit card.
Ethan Smith, CEO of Graphite, 18 years in SEOLenny's Podcast, Sep 2025Cited source. Not affiliated with The Super Panel.
Why this matters
A blunt reminder about what the "best of" lists assistants cite actually are in commercial niches. Knowing that is what stops you overpaying for a placement on one.

What not to bother with

  • Stuffing pages with phrases like "as an AI assistant" or other model-bait
  • Publishing volume for its own sake, which dilutes rather than accumulates
  • Paying for guaranteed citations, which nobody can supply
  • Chasing AI Overviews specifically, since they change too often to optimize against directly

The uncomfortable summary is that most of what works for AI citation is the same work that has always made a page good: be findable, be direct, be specific, be consistent.

This is not a separate discipline from search. It is the same work, why AI search optimization is sold as a tier rather than a package of its own, and why the writing that feeds it is priced alongside.

Keep reading

A tracked prompt set is the measurement this page rests on, and Google AI Overviews is a separate surface with its own behavior.

The rest of the writing covers costs and timelines, who should own the work is on in house versus outsourced, and whether the page can be read at all starts with the build.

Sources

Last updated 2026-09-07.

What people actually did when an AI summary appeared

A documented first-hand account, not a market pattern. Pew's KnowledgePanel Digital respondents install an application that records the URLs they visit; the finding is drawn from the recorded URL stream for 1-31 March 2025, not from a survey question, and the denominators are published with it.

Most claims about AI search rest on what people say they do. Pew recorded what a panel actually visited.

Across 68,879 Google searches by 900 US adults in March 2025, users clicked a traditional result on 8% of searches that produced an AI summary, against 15% of those that did not. Clicks on a link inside the summary happened on 1% of visits.

They also stopped browsing more often: 26% of pages carrying a summary ended the session, against 16% without one. It is one month, one US panel and Google only, and it measures clicks rather than whether the need was met.

The consequence is a measurement problem. Attribution is weak: this traffic often carries no referrer, so citation cannot be measured like clicks. A monthly prompt set tells you whether you appear.

It also rules some things out. Keyword stuffing does not work here any more than it does in traditional seo, and model version changes move results in ways nobody outside controls.

Any vendor offering promised citations is describing things that do not work. This runs alongside seo rather than instead of it, and calling it a waste of time is as unevidenced as guaranteeing it.

What this does not say: One month of data from one US panel, covering Google only. It measures clicks, not whether the user's need was met.

Pew Research Center, Do people click on links in Google AI summaries? - KnowledgePanel Digital browsing data, 68,879 searches by 900 US adults, March 2025 2025-07-22

Questions people ask

How do I get ChatGPT to recommend my business?

You cannot make it. You can make your pages crawlable, direct and specific, and you can measure citations on a fixed prompt set to see whether visibility changes. Anything stronger than that is a guess dressed as a method.

Is AI search optimization the same as SEO?

It overlaps heavily. The technical base, structure and specificity all serve both. The difference is measurement: a citation is binary where a ranking is positional.

Should I block AI crawlers?

A publishing business protecting its content may reasonably choose to. A business that wants to be cited cannot block the crawler and then expect the citation.

Does schema markup get me cited?

It does not force anything. It removes ambiguity about what your page is and who published it, which makes selection more likely rather than guaranteed.

How long does it take to see AI citations change?

Structural fixes can show up within a month or two on prompts where you were already close. Competitive prompts move slowly, and some never move because the model favors a source type you are not.

Do backlinks matter for AI citation?

Indirectly. Authority signals influence what surfaces in the retrieval step for many systems, so the same links that help search tend to help here too.

What is GEO or AEO?

Generative engine optimization and answer engine optimization: two labels for getting cited inside AI answers. The terms are used loosely and both describe the same work.

Is it worth investing in this yet?

It depends on whether your buyers ask models instead of searching. If your site is not indexed properly, this is premature and fixing indexing is the better spend.

Every price is on the page

Eight productized marketing services with published tiers, counted deliverables and printed exclusions. Compare them against any quote you already hold.

Entry point $300/mo. Cancel any time. No long contracts.

Still deciding?

Read how the process runs end to end: order, brief, production, delivery, report.

How it works
Read more posts