How to rank #1 on ChatGPT?

soldev's point above matches what I measured, so here's an n=1 data point for it.

I ran steps 1 and 3 as a fully automated system for two weeks — schema, sitemap, robots.txt, llms.txt, IndexNow, one long-form article per day built around real customer questions, with sourced statistics every few paragraphs. 18 articles across two sites.

Citations earned: zero. Revenue: $0.

The execution was essentially correct — auditing it afterward turned up only two dead links in my llms.txt and a robots.txt that Next.js was silently ignoring in favour of a generated one. Neither explains the result. What explains it is that the domains were three weeks old with nothing pointing at them. Steps 1 and 3 are cheap and automatable, which is exactly why they don't differentiate. Step 2 gates everything and is the one you can't script.

On your page-type question — the public numbers I could find (DeltaV Digital's 25k-citation analysis and Attrifast's 1.2M-citation study, both 2026) put articles at 23.7% of citations, listicles 19.6%, product pages 16.3%. But comparison pages had the highest per-retrieval rate, 1.87 vs a 1.29 average, on only 4.1% of volume. So comparison pages look like the most efficient shape and the least retrieved one. I'd also want that checked against a first-party dataset rather than aggregated studies.

The Reddit/Wikipedia citation split is the most useful concrete data point here, worth actually checking against, since it's a testable claim rather than a vague "be authoritative" tip.

Curious about step 5 in practice: with rankings this dependent on which sources currently get cited, how often does Alex actually detect a real shift versus noise in a single day's monitoring? That seems like the harder engineering problem hiding under "repeat continuously."

I knew ChatGpt had a ranking system

One measurement I would add is whether the person who lands from an AI answer recognises their actual job on the page. You can win a broad prompt and still attract people with no activating use case.

For Speechara, the higher-signal questions are task-shaped: “how do I keep the exact context from a customer call?” rather than “best AI tool.” We are trying to measure citations alongside first meaningful use, not citations alone. Otherwise ranking becomes another attractive top-of-funnel metric that does not tell you whether the recommendation was useful.

Used to spend weeks reaching out to publishers for backlinks . If your AI solves that, it's a game changer.

Your tool is amazing, it's on the point. I loved the frog in the right Nav Bar helping me xD

Thanks it's very interesting

Ranking for "best tool to build a business in 2026" is a vanity prompt: nobody with a credit card out types that. At SocialPost.ai the prompts that actually convert are narrow and problem-shaped ("how do I schedule posts across five client accounts"), and almost nobody competes for those. Are you tracking citation-to-signup by prompt, or just position?

Great breakdown. One thing I've noticed is that AI search engines don't just reward good SEO—they also favor content that directly answers user questions with clear explanations and real examples. We've been seeing this with our physics learning resources at, especially on topics like Projectile Motion, Newton's Laws, and the Ideal Gas Law. Well-structured educational content, strong internal linking, and relevant citations seem to make a noticeable difference over time. It'll be interesting to see how AI search continues to evolve.

I tried testing this method

Been testing this for clients. The playbook that's worked: (1) identify the 10-15 questions your customers actually ask ChatGPT (not what you think they ask), (2) create one page per question that answers it directly with data/citations, (3) get mentioned on 2-3 external sites in your space. We went from 0% AI visibility to showing up in ~12% of relevant prompts in 6 weeks for a client in insurance. Not magic — just systematic.

The report is the most useful thing in here, and I think it argues for step 4 harder than it does for steps 1-3. If Wikipedia is the top source for ChatGPT and Reddit leads for AI Overviews and Perplexity, then most of the leverage is in being mentioned somewhere the model already trusts, not in your own schema markup.

Worth pairing with the GEO paper out of Princeton and Georgia Tech (arXiv
2311.09735). They ran around 10k queries and found that adding authoritative citations and concrete statistics to a page lifted citation rate 22-41%, while keyword optimization and general fluency did close to nothing. That's a little awkward for the "clear keyword targeting, strong headings" part of step 3 — those are classic SEO fundamentals that don't appear to transfer.

Question on your dataset, since you actually have the citations: did you see any difference by page type? Definition pages vs comparison pages vs listicles. Where to get cited is one problem, but what shape the content has to be in is the one I'd pay for.

Great Post About Ai

The claim I would pressure-test is the #1 part, because model answers vary by user, memory, and session, so what you have is a favorable sample rather than a ranking. The number I care about is what share of new signups arrive with no referrer and name an AI tool in the how-did-you-hear-about-us field, which is the only AEO metric I have seen track revenue. If Alex reports that, you have a product. If it reports citation counts, you have a dashboard.

This is an interesting shift. I think AI visibility is becoming a new layer alongside traditional SEO rather than replacing it. I'm curious—which optimization had the biggest impact on your visibility across AI assistants? Was it structured content, authority signals, or something else?

The citation part is becoming the new backlink game. I’ve seen small tools get more visibility from being mentioned in the right places than from publishing 50 more blog posts. my website visiovnix.com got 1 million citaitons in 7 months

The Leapd example for "Best tool to build a business in 2026" is a great hook. I can see how systematic the process is, but as a solo founder, step 4 (getting cited by sources already ranking competitors) feels like the biggest time sink. Between building and supporting the product, there's barely room for outreach.

What I'd love to know: for a one-person team, how much time per week should realistically go into AI visibility vs. product work? Do you have a rough ratio?

This is becoming such a real concern for early-stage products — showing up in AI-generated answers might matter as much as SEO did. Have you found any specific tactics that actually moved the needle, or is it still mostly guesswork?

The piece I'd add to #3: it's not just "structured sections and direct answers" in the abstract, it's specifically whether the first 2-3 sentences of a section pre-answer the question on their own, without needing the reader to keep scrolling. I've been testing this on my own content and the pages that get pulled into AI answers are consistently the ones where you could delete everything after the second sentence and still have a complete, correct answer. Narrative or scene-setting openers basically never get cited even when the underlying info is good.

FAQPage schema has been the single highest-leverage structural change for me, more than internal linking or backlinks so far. It gives the model a pre-chunked, labeled unit that's trivial to lift wholesale.

Interesting topic !! when you say “rank #1 on ChatGPT,” do you mean getting recommended in answers, showing up in citations, or brand mentions? Would love to know which tactics actually moved the needle for you vs what just sounds good in theory.

Crawlability is step zero and a lot of sites block GPTBot without knowing it. We fixed ours and saw mentions pick up within a couple weeks. The piece I would add from testing on my own small product is that the prompt you type is not what gets retrieved. ChatGPT rewrites it into a few sub queries and fetches against those, so you can write the perfect page for "best tool to build a business in 2026" and it never gets pulled because it actually searched for "business builder pricing comparison."

Interesting topic. I think the difficult part is that there isn’t one fixed ranking like Google. Brand mentions, trustworthy content, third-party references, and clear product positioning probably all contribute over time. Would love to see what experiments actually moved the needle for you

did you have a backlinks from Replit? because I can see Replit logo there in the Chatgpt answer!

Really useful post. We worked with a company that had almost no online presence at all, and step 1 made the biggest difference for them. Clear text near the top explaining what they actually do mattered more than backlinks at that early stage. Curious if Alex changes its approach for industries where even the competitors barely have content, versus ones where everyone is already fighting for the same citations.

Great breakdown, Cyrus. The point about getting cited by the sources already influencing AI answers is especially valuable. Traditional SEO is still important, but AI visibility also depends on how clearly a brand is understood, mentioned, and supported across trusted sources. Consistently improving technical SEO, publishing helpful content, and building relevant authority seems like the strongest long-term approach.

The citation source point is what stood out most. Getting listed on directories and third party comparison pages has moved the needle more for my free tool site than anything I published on my own domain. One Reddit mention or a listing on a relevant directory sends stronger signals than ten blog posts on your own site because the model treats external mentions as validation rather than self promotion.

The chicken and egg problem someone mentioned in the comments is real though. The sources AI engines trust most are also the hardest to access when you are brand new.

good list. one thing on step 4 though, the prompt you type isn't really the thing that gets retrieved. chatgpt splits it out into a few rewritten sub queries and goes fetching against those instead, so you can have a page that's a perfect answer to "best tool to build a business in 2026" and it just never gets pulled, because what it actually went looking for was something like "business builder software pricing comparison."

took me way too long to figure that out tbh. now i just ask it what it searched rather than only reading what it answered, and the list of pages worth writing looks completely different.

does Alex sit at the prompt level or the fan out level? feels like that's where most of the "we did everything right and still didn't show up" stuff comes from.

Interesting breakdown. I especially agree that consistent third-party mentions may matter more than trying to optimize for one exact prompt. I’m building a small e-commerce tool now, so I’m curious: for a new product with almost no brand mentions yet, which first 2–3 channels would you prioritize to build credible visibility?

The one thing I'd push back on gently: ranking #1 for a single prompt is a sample size of one, and LLM answers are non-deterministic enough that the same query in a fresh logged-out session can return a completely different brand set. When we started measuring this seriously we switched to "share of voice across N runs" — fire each target prompt 15-20 times in clean sessions and record the percentage of runs you appear in. A lot of apparent #1 rankings turned out to be 25-30% appearance rates, and the sampling also made it obvious which citations were actually load-bearing versus incidental. Your step 4 matched our experience as the highest-leverage one: a mention inside someone else's "best X tools" comparison page got pulled into answers far more reliably than anything we published on our own domain, probably because the model treats third-party framing as evidence rather than a claim. Does Alex treat visibility as a binary rank position or does it sample each prompt repeatedly to build a frequency distribution, and are you seeing meaningful drift between logged-in accounts with memory enabled and clean sessions?

good breakdown, and the part most people miss is that AI-search ranking is driven less by your own on-page optimization and more by what OTHER sites say about you. these models synthesize from many sources, so being mentioned, reviewed, and cited across the web (reddit, comparison posts, directories, third-party articles) moves you more than tweaking your own schema. your own site being crawlable and clearly factual is table stakes, but the lift comes from off-site presence, the model has to see you referenced in multiple places to confidently recommend you. two practical adds: 1) answer-shaped content wins, a page that plainly states "X is a tool that does Y for Z" gets pulled into answers more than clever marketing copy, because the model can quote it cleanly. 2) it compounds with real usage, tools people actually discuss get cited more, so this isnt a one-time hack, its downstream of genuinely being talked about. curious how much of Alex's lift you attribute to on-site vs earning off-site mentions, because in my experience the off-site half is most of it.

Hey Cyrus, saw your post on Alex and AI search optimization. You've got the tech locked down, but every product needs a post-click conversion system. I build turn-key Discord architectures, Notion operational CRMs, and Whop sales funnels under Kinetic Digital Labs. Let's do a quick chat to see if my distribution/community stack fits your product launch.

The step most people skip is #4, and it's the one that actually moves the needle. When I looked at where the models were pulling answers in our category, it was almost never our own blog, it was third-party comparison pages and one Reddit thread. Cheapest win is getting listed on the five sources already being cited, not publishing more of your own content.

Great insights. I especially liked your point about validating ideas before scaling. Many founders focus on building first instead of understanding what users actually need.

Worth checking your robots.txt and WAF first — a lot of sites are invisible to ChatGPT because their own bot protection is blocking GPTBot and OAI-SearchBot, which are separate agents.

Also, models seem to quote Reddit threads and comparison posts far more than a company's own homepage — has anyone here asked the model to cite sources for your category?

I’d be careful calling one good answer a “ranking”, but it’s definitely encouraging! :-) We’re about to test this for our own site, and I think the real proof will be whether we keep showing up across different versions of the question and over time. Would be interesting to see the prompt set behind the result.

Ranking #1 on ChatGPT isn't about gaming the system—it's about creating content that's genuinely useful, accurate, and easy to understand. Focus on answering real user questions with clear structure, original insights, and up-to-date information. Build topical authority by consistently publishing high-quality content in your niche.

Also, make sure your website is technically optimized, earns mentions from trusted sources, and provides a great user experience. As AI-powered search grows, trust, relevance, and expertise matter far more than keyword stuffing, so prioritize helping users over trying to manipulate rankings.

Hi, I am new in this arena so was wondering if i follow all the steps to increase GEO of my website, how would i know if its working? what's the right approach for testing the effectiveness?

Two things I would add to step 4, both learned by getting them wrong.

Fix one exact brand string before the first measurement and keep it fixed. I track a fintech whose name is also an ordinary word, and when the tracked string fell back to the bare brand the mention count went from 488 to 159,512, because a large cosmetics brand shares that name. A jump like that reads as growth on a chart and carries no information. The trend line is comparable only if the string stays identical between runs, which matters for any company named after a common word.

Track sentiment alongside citations. I found sites with strong citation rates where the pages being cited carried negative reviews, so the models were reading about them and repeating the wrong thing. Citation count tells you that you are inside the answer, not how you appear in it.

Separately, I built my own tracker for prompts, mentions and citations, and I am looking for people who work in this area to look at it and say what is missing. If you have spent real time on AI visibility measurement, I would appreciate a reply.

Strong framework. The distinction between publishing content and becoming a trusted citation source is especially important.

At 6senseHQ, we are seeing that schema, crawlability and structured content create the foundation, but third-party mentions and citation consistency often determine whether a brand actually appears in AI-generated answers.

Which metric do you consider the most reliable for measuring progress: prompt-level visibility, citation frequency, referral traffic from AI platforms or eventual conversions?

One technical detail I’d add: crawlability is not only about robots.txt and schema. For data-heavy products, the key explanation should exist in the initial HTML, not only after JavaScript fetches the data.

I’m building a WordPress dashboard with live market-data fragments, and I’ve found it useful to keep the first server-rendered response meaningful, then use JavaScript for charts and partial refreshes. This also gives the page a stable description when an API is temporarily unavailable.

Does Alex distinguish the static, crawlable product explanation from dynamic content when auditing a site?

The crawlability point is underrated — we found our robots.txt had been silently blocking GPTBot for months, and branded mentions in ChatGPT answers picked up within a few weeks of fixing it. Curious how Alex sources the editorial backlinks though; that feels like the hardest part to scale without it turning into spammy outreach.

step 5 is the part most people skip. rankings decay fast once you stop publishing. been seeing the same thing on the competitor side in bunzee the gap list looks totally different month to month. how often does alex re-audit?

I faced a similar issue while building a gaming tool. The most important part was keeping the interface simple and making the value comparison easy to understand on mobile.

I recently built an MM2 trade checker where players can compare Murder Mystery 2 items before accepting a trade. I am still improving the item data, speed, and overall user experience, so feedback from other builders would be really helpful.

"Great write-up. This proves that LLM recommendations aren't just random luck. Getting Leapd to show up first means your data graph is highly structured and well-indexed by OpenAI. Are you planning to track how much conversion or direct traffic this specific ChatGPT recommendation is bringing you?"

I agree that the focus should be on building authority instead of trying to game AI search. It seems like AI visibility probably compounds over time. It also seems like the businesses creating genuinely valuable content will have a much bigger advantage over the long run than those chasing quick wins.

The directory and Reddit point resonates a lot with what I've been running into this week launching my first product. Turns out getting into those citation sources is its own gate: several directories require the account or the domain to have some minimum age before they'll even accept a submission, and Reddit gates new accounts by karma before certain subs will let a post through. So there's a chicken-and-egg problem underneath all of this — the sources AI engines trust most are exactly the ones hardest to access when you're brand new, regardless of how good the content or product actually is.

Makes me think "AI visibility" for an early-stage project is less about optimizing pages and more about just surviving long enough (in account-age, karma, backlink terms) to become eligible for the channels that actually get cited.

Great breakdown. I think documentation is an underrated piece of AI visibility. Well-structured docs and tutorials often end up being cited because they're easier for AI systems to understand than marketing pages. It feels like teaching your product is becoming just as important as promoting it.

Curious what your top signal ended up being. From what I've seen shipping a small set of browser-based tools: the pages ChatGPT cites most are the ones that answer one specific question in the first paragraph with a clean number or fact. Not the 2,000-word "ultimate guide" — those get ignored. The "what's the CBM of a 20ft container" type page wins.

Schema markup (FAQ / HowTo) helps, but only after the content is actually structured that way naturally. I tried forcing it once and the citations dropped — felt like ChatGPT treats over-marked-up content as less trustworthy, not more.

When I tried to get there was different results on different account,

Wow, this is news to me, I didn’t know you could influence whether your website appears in ChatGPT’s responses. I’ll definitely use these tips. Thanks!

just built my first business .. insane

One thing I've noticed while building my own SaaS is that AI visibility feels much harder to measure than SEO. With Google you can track rankings, but with ChatGPT it's difficult to know whether you're actually improving or just getting lucky. Curious if you've found a reliable way to measure progress over time.

The citation source breakdown is the most useful part here: Wikipedia for ChatGPT, Reddit for Perplexity and AI Overviews. We saw the same at SocialPost.ai, a couple of Reddit mentions moved our AI visibility more than months of schema cleanup. Do prompt rankings hold once a competitor lands a fresh comparison article, or do they flip within days?

The crawl lag point is real. I've been experimenting with getting my Flutter apps surfaced in AI answers, and the gap between publishing good content and seeing it cited feels like weeks, not days. One thing that helped was making sure the app's landing page answers the exact question in plain text near the top ? not buried in a hero section or behind a CTA. AI engines seem to reward the first clear paragraph more than flashy design. Would be curious if Alex accounts for answer positioning in page structure.

I'm studying also this things, GEO is becoming really important but maybe is still not clear how to get the key of success. I think there are several combinations to rank and get suggested from AI, crawler work differently (gemini, openai etc..)

The crawl lag point is real. I've been experimenting with getting my Flutter apps surfaced in AI answers, and the gap between publishing good content and seeing it cited feels like weeks, not days. One thing that helped was making sure the app's landing page answers the exact question in plain text near the top ? not buried in a hero section or behind a CTA. AI engines seem to reward the first clear paragraph more than flashy design. Would be curious if Alex accounts for answer positioning in page structure.

Solid breakdown — 1 through 5 are the actual fundamentals and it's useful to see them written out plainly rather than as a black box.

The thing I'd add: a single "we ranked #1" screenshot is a sample of one, and these answers are non-deterministic. The metric that matters isn't "did we appear" but "what share of runs cite us" for the same prompt over time — one flattering result and a stable citation rate across 50 runs are very different claims.

Genuinely curious how Alex handles that — does it track citation rate across repeated queries, or point-in-time position? That variance piece feels like where the real proof lives.

Interesting perspective. One thing I've noticed is that it's less about "ranking #1" and more about becoming a consistently citable source. Clear answers, original insights, and authority across multiple platforms seem to matter more than traditional SEO tactics.

The thing that surprised me when I actually measured this: the traffic is real, the volume is small, and the intent is very high.

Over 90 days ChatGPT sent us 16 visitors and 30 pageviews. Tiny. But in the period I checked it was one of only two channels that produced any signups at all, while classic social produced zero. So the ROI per visitor is nothing like the ROI per visitor from a feed.

What seemed to move it was not a schema tweak. It was having the direct answer to the exact question inside the first two sentences, then a real number with a source attached. The pages that got cited were the ones where the answer was extractable without reading the whole article.

One practical thing: split it out from Google in your analytics rather than leaving it blended under organic. Blended, 16 visitors disappears into the noise and you conclude it does nothing.

This is great. Thank you for contents like this 🫡

The backlink point tracks with what I've seen, but I'd push back a little on the "get cited by the sources already ranking your competitors" step being purely a content quality problem. A lot of what shows up in AI Overviews for B2B-ish queries is old Reddit threads and Quora answers that never got updated, not necessarily the best current source, just the one the model already trusts and has indexed. There's a lag between "good content exists" and "the model actually surfaces it" that seems more about crawl and citation timing than about the content itself.

Curious whether Alex accounts for that lag somehow or just flags the gap and you're stuck waiting for the next crawl cycle. That's been the more frustrating part on my end than the content or backlink work itself.

Great breakdown. I’ve been testing AI workflows for content creation, and one thing I’ve learned is that distribution matters much more than prompt quality. Thanks for sharing your experience.

Building on mocktomer’s point that none of this shows up in analytics, because I think it is the most solvable part and the least discussed: you cannot measure citations in analytics, but you can measure them in server logs.

Disclosure, I work on growth at Ojin and we track this for our own site.

Assistant fetchers identify themselves by user agent, and the distinction that usually gets collapsed is between training crawlers and live retrieval fetchers. GPTBot is largely corpus collection. OAI-SearchBot and PerplexityBot fire when a person has actually asked something and the engine goes out to retrieve. So a hit from the second group is a leading indicator that a specific page entered the retrieval path for a live query, often before any citation is visible to you. That signal is free, already sitting in your logs, and it does not depend on anyone clicking through, which is exactly the gap mocktomer described.

It also makes a strategic choice available that has not come up here: those user agents can be handled separately in robots.txt. You can decline the training crawlers while staying fully available to the retrieval fetchers, so you remain in the answers without feeding the next model. Whether you want that is a judgment call, but most people do not realise it is a separable decision at all.

And on Korelyy’s question about whether FAQ or HowTo schema actually moves citations or whether the content structure is doing the work: worth knowing Google deprecated FAQ rich results for most sites back in 2023, so that markup no longer earns a SERP feature. What it still does is force you to write in question and direct-answer pairs, which is the shape that gets extracted. So the honest answer is probably that the structure is doing the work, and the schema is a useful discipline for producing it rather than a ranking input in its own right.

​That shift toward how AI search engines choose their sources is fascinating. As mentioned here, there is a massive difference between what triggers a citation in ChatGPT versus Perplexity or Google AI Overviews.
​I am currently building out an independent e-commerce and software ecosystem where we are looking quite a bit at how to automate optimization and skip the manual grunt work. It's super interesting to see how tools like Alex are trying to solve it.
​Anyone else here experimenting with structuring their site specifically for AEO (AI Engine Optimization) yet, or are you still sticking purely to traditional SEO?

This was really helpful guide which triggered me to try the service. Nice to see people sharing their findings in details

Been deep in this exact rabbit hole — I test how AI assistants see websites, and I ran the check on my own site first: scored 33/100, visibility literally 0. What actually moved the needle in testing: (1) machine-readable product data — schemadotorg Product/Offer markup, real prices in HTML instead of baked into images; (2) an llms.txt plus clean category pages crawlers can parse; (3) making sure key pages render without JavaScript — most assistant fetchers don't execute it; (4) being present in places assistants already trust: directories, comparison pages, and review sites get cited constantly in their answers. The uncomfortable part: none of this shows up in analytics, because the shopper who asked ChatGPT never visits your site at all. You lose them before the click exists.

The schema + crawlability piece is underrated - most people jump straight to backlinks and skip step 1 entirely. One thing I have noticed running a tools site: AI engines seem to cite pages that answer one specific question cleanly far more than long "ultimate guide" pages, even when the guide has more authority behind it. Direct-answer blocks and clear question-style H2s get pulled into ChatGPT/Perplexity summaries much more often for me. Have you seen structured data like FAQ or HowTo schema actually move citations, or is it mostly the content structure doing the work?

the Reddit signal is the one most people underestimate. when we looked at where ChatGPT was pulling from for AI-related queries while building aisa.to, community threads and niche discussion posts showed up way more than traditional blog content. schema and structured data help with crawlability, but being genuinely cited in conversations where people are already asking questions is what actually moves rankings.

It's interesting to see how your platform helps for AI visibility search. gonna try it!

It's interesting to see some much of the fundamentals between SEO and AEO is the same. I like the breakdown in the report . Although some parts are quite different that what I expected.

getting blog mention for my business was a huge pain, if your AI finds publishers and reach out to get backlinks that is a huge win .. let me try the free visibility

thanks for sharing - I wonder how long it takes to get similar results

I like the breakdown and the structure make sense, but how the backlink works, like your AI agents find publishers that would give my business a backlink or there is a directory we pick from?

I think the biggest advantage of this system is the backlink building and fresh content - website audit is foundational so if someone doest do that thy probably not even selling anything ...

def love the itemized plan - it was under the impression wikipedia is not a reliable source , I am SURPRISED to see it is the top source for chatgpt!!!! how many prompts did you analysis for this?