Getting Found in Google and Cited by AI: I Checked the Numbers

Someone sent me a neat little summary on how to get cited by ChatGPT and Google's AI answers. It was credited to Neil Patel and NP Digital. Ten page types, each with a percentage. Lift numbers for YouTube, Reddit and review sites. A quote or two.

It looked like a plan you could build a website from. Before I built anything on it, I checked every number against the original sources. Some held up. Some were changed along the way. A few I couldn't find anywhere.

I'm Damon Nelson. I've been a marketer and developer for more than 30 years, and I'd rather build on numbers I've checked than numbers that sound good. Here's what holds, what doesn't, and the plan I'd follow for a new site. The research was checked on November 1, 2026. There's a free PDF of the whole report at the bottom.

The short version

AI answer engines mostly pull from the same pages search engines already trust. So good SEO is still the foundation. On top of that, three things move the needle in the evidence I could check:

1. Publish facts nobody else has. Your own data, tests and prices.
2. Put the answer first. Write each page so the answer sits in the first few sentences under a clear heading, with tables for anything that has numbers.
3. Get talked about on other sites. NP Digital found that on buying-type questions, 82% of AI citations go to third-party sites and only 3% go to the brand's own pages.

Page format matters less than people claim. In that same study of 9 million AI answers, 64% of citations came from ordinary pages.

How I checked it

I went back to neilpatel.com, Search Engine Land, Google's own documentation and independent research. Every claim got one of three labels:

• Verified: found in the original source, with a link.
• Partly verified: the source exists, but the summary changed the wording, the meaning or what was measured.
• Unverified: I couldn't find it. These only show up in the "claims I couldn't confirm" table near the end.

One thing worth knowing up front. NP Digital's page-type rankings come from surveys of marketers. They show what marketers said they saw, not measured AI citations. NP Digital also sells AI search services and owns AnswerThePublic and Ubersuggest. Its 9-million-answer study is closer to a real measurement, but the full method isn't published. None of that makes the data wrong. It just tells you how much weight each number can carry.

Part 1. The summary, checked line by line

NP Digital's own studies point in different directions

Comparison pages land second in one survey and last in the other. That's the tell. No format is a magic key. Formats help an engine lift your answer. Depth, original facts and outside mentions decide whether you get picked.

NO FORMAT IS A MAGIC KEY. ORIGINAL FACTS ARE.

Part 2. What it takes, start to finish

Step 1. The technical foundation (week 1)

This part gets overcomplicated more than any other. Google says it plainly: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." A page has to be "indexed and eligible to be shown in Google Search with a snippet." (Google Search Central, verified.) So the first job is plain old SEO housekeeping.

Here's a robots.txt for a site that wants to be found everywhere:

# Example robots.txt for a site that wants to be found everywhere
User-agent: *
Allow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

# Training bots: your choice
User-agent: GPTBot
Allow: /

User-agent: Google-Extended
Allow: /

Sitemap: https://www.example.com/sitemap.xml

Step 2. Tell the machines who you are

AI tools and Google both need to know who you are before they'll recommend you. Google calls this E-E-A-T: experience, expertise, authoritativeness and trust.

Use the exact same business name, address, phone and description everywhere. Build an About page and named author pages with real bios. Add Organization schema whose sameAs links point to your LinkedIn, YouTube, Google Business Profile, X and any directory listings. If you serve customers locally, claim and complete your Google Business Profile.

Step 3. Write pages an engine can quote

• Answer first. Put the direct answer in the first 40 to 60 words under each heading, so the engine can lift it without rewriting.
• One idea per heading. Write headings as the questions people actually ask.
• Use tables for anything with numbers. Prices, specs, comparisons, results.
• Add evidence. The GEO paper from researchers at Princeton and Georgia Tech (Aggarwal et al., KDD 2024, verified) found that adding statistics, quotations and citations to sources raised visibility in AI answers by up to about 40% in controlled tests, and by up to 37% on Perplexity.
• Be the source. Original data, your own tests and real prices are the content AI can't write without you.
• Make ordinary pages thorough. Most citations come from regular pages (64%), so depth on your core service pages matters as much as special formats.
• Date and update. Show a "last updated" date and refresh your key pages.

Step 4. Get talked about somewhere else

On buying questions, the evidence says other sites speak for you far more often than your own site does. That's Part 4.

Part 3. The extra pages worth building, ranked

These are ranked by NP Digital's 500-marketer survey score. Remember, that's a marketer rating, not a measured citation rate. Each one comes with a structure you can hand to a writer. I left generic blog posts (25%) and video pages (2%) out as page types to build.

1. Original research and data pages (survey score: 82%)

This page holds facts that exist nowhere else, so an AI engine has to point to you to use them. It also lines up with the Princeton GEO paper, where adding statistics, quotations and cited sources raised visibility by up to about 40%.

H1: [Topic] Benchmark [Year]: [Headline finding in plain words]
First 40-60 words: the single biggest finding with its number, sample size and date.
H2: Key findings (5-8 bullets, one number each)
H2: Methodology (who, how many, when, how collected, limits)
H2: Full results (HTML tables, one per question)
H2: What this means for [audience]
H2: How to cite this data (suggested citation line + permalink)
Schema: Article or Dataset; author as Person with sameAs links

2. Comparison and "vs" pages (survey score: 76%)

Buyers ask AI "A vs B" and "alternatives to A." NP Digital says comparisons only stand out when they're built on your own testing. A rewrite of feature lists performs like a generic blog post. And remember the other NP Digital survey put comparison and alternatives pages last (2.5%).

H1: [A] vs [B]: Which Is Better for [Use Case]?
First 40-60 words: the verdict. "Choose A if... Choose B if..."
H2: Quick comparison (HTML table: price, key features, limits, best for)
H2: How we tested (your criteria, dates)
H2: One H2 per deciding factor, with a winner line each
H2: Who should pick A / who should pick B
H2: FAQ (3-5 real questions)
Schema: Article; Product with offers only if you sell one of them

3. Rankings and best-of lists (survey score: 57%)

These answer "what are the best X for Y" prompts head on. In the 1,000-marketer survey, listicles earned the largest share of reported citations (21.87%). A ranked list names competitors by design, so be fair.

H1: The [N] Best [Things] for [Audience] in [Year]
First 40-60 words: the top pick and why, plus how you chose.
H2: Summary table (rank, name, best for, price, one-line verdict)
H2: How we ranked them (criteria)
H2 per item: "1. [Name]: best for [segment]" then pros, cons, price, who should skip it
H2: FAQ
Schema: Article plus ItemList

4. FAQ hubs (survey score: 41%)

Question-and-answer text maps neatly onto how people prompt. One catch: since August 2023, Google shows FAQ rich results mainly for government and health sites. FAQPage markup still helps machines read the page, but it rarely produces the visual rich result.

H1: [Topic] FAQ
Intro: one sentence on what the page covers.
H2 per question, phrased the way customers type it ("How much does X cost?")
First sentence under each: the full answer (40-60 words), self-contained
Then: detail, a number, a link to the deeper page
Schema: FAQPage, matching the visible text word for word

5. Step-by-step how-to guides (survey score: 39%)

Clear procedures are easy to lift into an answer. But Patel's own 1,000-marketer survey ranked how-to 8th of 9 (6.22%), so treat it as useful, not a top lever. Google removed HowTo rich results in 2023.

H1: How to [Do Task] ([N] Steps)
First 40-60 words: what you will end with, time needed, tools needed.
H2: What you need (bullets)
H2: Step 1: [Verb + object] ... one H2 per step, imperative sentences, a screenshot if it helps
H2: Common mistakes
H2: FAQ
Schema: Article (HowTo markup is harmless but no longer shows in Google)

6. Community, forum and help center pages (survey score: 28%)

In the NP Digital survey this means showing up in forums, Reddit, LinkedIn and Quora, not only hosting a forum. On your own site, a help center full of solved questions plays the same role.

Help article H1: the exact problem in the user's words ("Error X when doing Y")
First 40-60 words: the fix.
H2: Why it happens
H2: Step-by-step fix
H2: Related questions (links)
Schema: Article or QAPage (for real user Q&A with answers)

7. Definition and concept explainers (survey score: 22%)

Low on its own, because AI already knows most definitions. It's worth building only when you add a number, an example or a viewpoint AI can't produce without you.

H1: What Is [Term]?
First sentence: "[Term] is ..." in 40-60 words.
H2: How it works
H2: Example (your own, with numbers)
H2: [Term] vs [related term] (small table)
H2: FAQ
Schema: Article; DefinedTerm is optional

8. Opinion, thought leadership and case studies (survey score: 16%)

NP Digital's 16% is for opinion and thought leadership. Case studies weren't a separate item. A case study with real before-and-after numbers works more like original research, so build it that way.

H1: How [Client] [Got Result] in [Time]
First 40-60 words: client, problem, result with numbers.
H2: Results at a glance (table: metric, before, after)
H2: The starting point
H2: What we did (numbered)
H2: What did not work
H2: Quote from the named client (with permission)
Schema: Article; Organization for client and author

9. Product and service pages (survey score: 14%)

Your own pages rarely get cited on buying prompts (3% in the 9-million-answer data). But when they do, NP Digital saw about a 5x visibility lift. Make the facts easy to quote: specs, prices, integrations.

H1: [Product]: [What it does] for [who]
First 40-60 words: what it is, who it is for, starting price.
H2: Pricing (HTML table, real numbers)
H2: Specs / what is included (table)
H2: Works with (integrations list)
H2: Who it is not for
H2: FAQ
Schema: Product with Offer, or Service; Organization

About, leadership and trust pages (not in the survey)

I couldn't find the "34% attribution lift" from the summary anywhere. Build this page anyway. It tells Google and AI who you are, and every other page leans on that.

H1: About [Brand]
First 40-60 words: what the business does, for whom, where, since when.
H2: Founder / team (named people, real bios, photos)
H2: Credentials (licenses, awards, press, memberships, with links)
H2: Contact and location
Schema: Organization (or LocalBusiness) with sameAs links to LinkedIn, YouTube, Google Business Profile, Wikipedia if one exists; Person for founders

Part 4. The other sites AI leans on

This is the part most people skip, and it's the part that does the most. These numbers come from NP Digital's study of 9 million AI answers across 9 platforms and 400+ enterprise brands (neilpatel.com, Search Engine Land). "Lift" means how much more visible a brand was in AI answers when it was cited from that source.

Also worth doing, based on how these answers get built: get listed in roundups and "best of" articles on other sites, get quoted in industry press, and keep your directory listings consistent. Patel's advice from the 9-million-answer study is to aim digital PR at buying-stage topics like pricing comparisons, category roundups and "best for" lists.

Part 5. The 30, 60 and 90 day plan

Part 6. The checklist

Claims I couldn't confirm

What I'd take from all this

The summary I was sent wasn't useless. Most of its numbers are real. They just meant less than they looked like they meant, and a few came from nowhere I could find.

The plan that's left is less flashy and a lot more solid. Get indexed. Answer first. Publish something only you can publish. Get other sites talking about you. None of that needs a trick, and none of it goes stale when the next viral summary shows up.

That's also why this page exists. It's my own check of the numbers, with every source linked, which is exactly the kind of page the research says gets cited.

Build it with us

At GeekOutFridays we spend 30 minutes on strategy like this and 60 minutes building one system with it. The mastermind meets live every other Friday.

Join the GeekOutFridays mastermind group. Sign up here.

Sources

1. NP Digital, "Content Types That Perform Best in AI Search," May 2026 (500 marketers)
2. NP Digital, "Which On-Site Content Types Earn the Most AI Citations?," Aug 2026 (1,000 marketers)
3. Neil Patel, "AI Visibility at Scale: Learning From 9M Prompts and 400+ Enterprise Brands"
4. Search Engine Land, "What 9 million AI answers reveal about search visibility"
5. NP Digital, "Reddit's AI Visibility Drop-Off," Oct 2025
6. Google Search Central, "AI features and your website"
7. Google Search Central blog, changes to HowTo and FAQ rich results, Aug 2023
8. Google, common crawlers (Google-Extended)
9. OpenAI, overview of crawlers (OAI-SearchBot, GPTBot)
10. Perplexity, crawlers (PerplexityBot)
11. llms.txt proposal
12. Aggarwal et al., "GEO: Generative Engine Optimization," KDD 2024
13. Search Engine Land, NP Digital acquires AnswerThePublic, May 31, 2022