The MBC Group
HOMEAGENT TEAMSSEO AgentGoogle Business Profile AgentGoogle Ads AgentMeta Ads AgentSocial Media AgentAI Search AgentLead Conversion AgentVoice & Chat AgentsWorkflow AutomationBUILDWebsite DesignApp DesignBrandingAIDEN OSPRICINGCUSTOMERSABOUTBLOG
LET'S TALK »
MBC / Group / Blog / SEO & Traffic
SEO & Traffic 12 min read Sep 1, 2026
Search “how to rank in chatgpt” and Google returns twenty results across sixteen websites. On 1 September 2026 we opened every one of them. Eleven are written guides and we read all eleven in full; the other nine are YouTube videos, LinkedIn posts, a Reddit thread, a Quora page and a Medium post.
Then we ran the test none of them runs on itself. Every one of these pages is teaching you how to get ChatGPT to read your website, so we checked whether ChatGPT is allowed to read theirs. Five of the twenty sit on sites whose robots.txt tells OpenAI’s crawler, in writing, to stay out. And of the eleven guides, six tell you to display a visible “last updated” date — while none of the eleven displays one.
Matthew Montez
Founder · MBC Group
Share
On this page
00 Key Takeaways 01 What we did 02 Five pages block ChatGPT’s crawler 03 OpenAI runs three crawlers 04 Nobody shows a last-updated date 05 25 mentions of schema, no schema 06 What page one gets right 07 What we would do first 08 Method, and what this does not prove 09 Common questions
Related
→ AI Automation → AI Agents → AI-Enhanced SEO → Google Ads → Free Marketing Audit
Free audit
See where AI moves the needle fastest.
Talk to Aiden →
A business owner hears that customers have started asking ChatGPT which company to hire, and does the obvious thing: opens Google and searches for how to rank in ChatGPT. On 1 September 2026 we ran that exact search from a United States location and read everything that came back — twenty results, twenty distinct pages, sixteen distinct websites.
Nine of the twenty are not articles at all. Four are YouTube videos, two are LinkedIn posts by the same author, one is a Reddit thread, one is a Quora page and one is a Medium post. Eleven are written guides on a company website, and we read all eleven of those in full.
Then we ran a test none of the guides runs on itself. Every one of these pages is trying to teach you how to get ChatGPT to read and cite your website. So we checked whether ChatGPT is allowed to read theirs. Five of the twenty pages Google ranks for this question sit on websites whose robots.txt file tells OpenAI’s crawler, in writing, not to come in.
00 — Key Takeaways
Five of the twenty ranking pages block GPTBot, OpenAI’s crawler, in robots.txt: both LinkedIn posts, the Reddit thread, the Quora page and the Medium post.
Reddit’s entire robots.txt is two lines long, and it blocks every crawler ever written — Google’s included. It is the fourth result for this search.
Quora names GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot and ClaudeBot one at a time and gives each of them Disallow: / — while leaving Googlebot free to read the page.
Six of the eleven guides tell you to display a visible “last updated” date. Zero of the eleven display one.
One guide mentions schema markup 25 times, includes a section headed “Use Schema Markup”, and ships no structured data at all — checked in a real browser after all its JavaScript had run.
Five of the eleven guides never once mention GPTBot, OAI-SearchBot, ChatGPT-User, robots.txt or llms.txt. The highest-ranking article of the eleven does not contain the word “crawl” anywhere.
We took the top twenty organic results for “how to rank in chatgpt” and checked them for duplicates first, because Google frequently returns the same URL twice and it quietly inflates any count built on top of it. This time it did not happen: twenty results, twenty distinct URLs. Two of them are different posts by the same author on LinkedIn, and four are different videos on YouTube, which is why sixteen websites produce twenty results.
For each of the eleven readable guides we pulled the page, stripped the navigation, header, footer and sidebars, and counted only the words in the article itself. That step matters: a menu bar that says “AI SEO” is not the same as advice, and counting raw page text instead of article text is how you end up reporting things that are not there.
Then the crawler test. For each of the twenty ranking URLs we fetched that site’s live robots.txt and evaluated the exact URL against it with a full matcher — wildcards, end-of-path anchors, longest-matching rule wins, and Allow beating Disallow on a tie — rather than skimming the file by eye. We checked eight crawlers: GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot, Google-Extended, Googlebot and Bingbot.
Here is the result for the crawler that decides whether your page can become part of what ChatGPT knows.
Ranking page
Position
GPTBot
OAI-SearchBot
ChatGPT-User
LinkedIn post
Blocked
Allowed
Reddit thread
Quora page
10
Medium post
19
The other 15 results
The eleven written guides all pass. Every one of them lets all three OpenAI crawlers in. The blocking is entirely on the social and community platforms — which is awkward, because those platforms are exactly where several of the guides tell you to go and build a presence.
Reddit is the sharpest case. Its robots.txt is not a long file with carefully negotiated exceptions. Stripped of comments it is two lines: a single user-agent group matching everything, and Disallow: / underneath it. There is one user-agent line in the whole file. Reddit tells GPTBot no, and it tells Googlebot exactly the same thing. Reddit still appears in both because those are commercial licensing arrangements, not permissions granted in a text file — which is a route that is available to Reddit and not to you.
Quora is more pointed. Its robots.txt opens with a notice stating that all crawlers are prohibited from using Quora content to train AI models without a contractual agreement, and then names the AI crawlers individually and blocks each one outright: GPTBot, OAI-SearchBot, PerplexityBot, ChatGPT-User, ClaudeBot. Googlebot, further up the same file, gets an ordinary list of excluded paths and is otherwise welcome. Quora ranks tenth on Google for a question about ranking in ChatGPT while explicitly refusing ChatGPT.
And the guide sitting at position eleven, from Ahrefs, contains this instruction: engage authentically on Reddit, which it calls ChatGPT’s most-cited domain. The advice may well be right about where citations come from. It is worth knowing that the destination it points at is the one page on this entire search result that refuses every crawler on the internet.
Most robots.txt files we see treat “AI bots” as one category to be allowed or denied together. OpenAI actually operates three, and they do different jobs. Blocking the wrong one costs you something different.
LinkedIn has clearly worked this out. It gives GPTBot, ChatGPT-User, ClaudeBot, PerplexityBot and Google-Extended a flat Disallow: / — and gives OAI-SearchBot nothing but a short list of excluded paths, which leaves the ranking posts readable. Training crawler out, search crawler in. Whatever you think of the position, it is a considered one, and it is the single most sophisticated piece of AI-crawler policy on the entire results page. Medium takes a similar line by accident of omission: it blocks GPTBot and ClaudeBot by name and never mentions OAI-SearchBot, which therefore gets in.
The gap, in one line
Twenty pages rank on Google for how to rank in ChatGPT. Five of them are pages ChatGPT is not allowed to read.
MBC Group, 1 September 2026
Freshness is the second most common piece of advice on this results page, and it comes with a specific instruction attached. Writesonic puts it plainly: always add a visible “Last updated” date, because it signals freshness to both users and AI systems. StudioHawk says to add “last updated” timestamps visibly on-page. web99 suggests the exact wording to use. Singularity Digital, Wellows and omnius all say some version of the same thing.
We searched every one of the eleven articles for a visible last-updated stamp. We found seven matches, and every single one of them was the article telling the reader to add one. Not one of the eleven guides displays a last-updated date of its own. Six give the instruction; zero follow it.
That would be a small point if the pages had not been updated. They have been — their own structured data says so, and the gap between what they show a reader and what they quietly record is sometimes months.
Guide
Date shown to the reader
Last modified, per its own schema
Ahrefs
13 March 2026
26 August 2026
StudioHawk
23 March 2026
18 August 2026
Writesonic
2 September 2025
28 January 2026
Power Digital
no date shown
10 February 2026
Neil Patel
11 May 2026
LLMrefs
18 January 2026
Ahrefs is the clearest example. A reader arriving today sees 13 March 2026 and reasonably concludes the page is nearly six months old. Ahrefs’ own markup says the page was last modified on 26 August 2026 — six days before we checked. The work was done. The reader is simply not told. StudioHawk, which is the guide that tells you to put the timestamp on the page, shows a March date on a page it modified in August.
Two more worth naming. web99’s guide is titled “How to Rank on ChatGPT in 2025”; it was published on 5 August 2025, its structured data records no modification since, and it still ranks fourteenth in September 2026. And omnius displays 19 August 2026 next to its author while its structured data says the post was published on 17 November 2025 and carries no modification date at all — so the date the reader sees and the date the machine reads do not agree.
Seven of the eleven guides recommend adding structured data. Ten of the eleven actually have some. The eleventh is TripleDart, which ranks eighteenth.
TripleDart’s article uses the word schema twenty-five times. It contains a numbered section headed “Use Schema Markup”, and tells the reader that schema helps AI models understand what a page is about. The page itself carries no structured data whatsoever — not an Article, not a FAQPage, not an Organization. We checked this twice, because a raw page fetch is not proof: plenty of sites inject their markup with JavaScript, and a naive check would have called that a false negative. We loaded the page in a real Chrome browser, let every script finish, and read the structured data out of the finished document. There was none.
For what it is worth, the FAQ question format that AI answers lift from most readily is present on only five of the eleven — Neil Patel, omnius, Writesonic, Wellows and LLMrefs. Six of the guides advising you to be quotable have not made themselves quotable.
This page has real strengths and it would be dishonest to skip them. Ten of the eleven guides ship structured data. Nine of the sixteen websites publish an llms.txt file, which is a higher adoption rate than we have found in any other category we have studied. And on the single most important technical point, page one is correct and close to unanimous: ChatGPT’s live retrieval leans on Bing’s index, so a site that Bing has never crawled has a problem no amount of content will fix. StudioHawk, omnius, Writesonic, Wellows and TripleDart all say so explicitly.
The advice about answering questions directly, keeping paragraphs short, using clear headings and earning mentions on third-party sites is also sound. We would not argue with any of it. Our finding is narrower and more awkward than “this advice is wrong”: it is that a majority of these pages do not do the specific, checkable things they instruct you to do, and that five of the twenty results are on platforms that have told OpenAI to stay out.
If you want to be cited by ChatGPT, the order of operations matters more than the length of the list. This is the sequence we run, and the first three items cost nothing but attention.
If you would rather not do any of that yourself, this is the work our AI search optimization service does, and the same audit that produced this study is what we run on a client site in week one.
The search was run once, on 1 September 2026, from a United States location, and search results move. Anyone who reruns it later will see a different page, which is why every specific claim above names the page it came from. All twenty URLs were checked for duplicates before anything was counted. The eleven readable guides were fetched directly and read in full; word counts and phrase counts were taken after navigation, headers, footers and sidebars were removed.
Two limits worth stating. First, robots.txt is a request, not a wall — it records what a site has asked crawlers to do, which is what we set out to measure, but it does not prove what any crawler did. Second, a robots.txt file can change any day; ours describes 1 September 2026. We have not tested whether any of these pages is in fact cited by ChatGPT, and we make no claim either way. What we tested is narrower and entirely checkable: what each site’s own file says, and whether each guide does the specific things it tells you to do.
We ran the same style of read on local SEO for small business and on who actually owns page one in Denver, if you want to see the method applied to different ground.
Q How do you rank in ChatGPT?
Q What is GPTBot and should I allow it?
Q How do I check whether ChatGPT can read my website?
Q Does ChatGPT use Google or Bing?
Q Does schema markup help you rank in ChatGPT?
Q Is llms.txt worth adding?
Q How often should I update a page for AI search?
Q Who did this research?
Find out whether AI search can see your business
MBC Group runs the same crawler, index and citation audit that produced this study. It takes a week and it usually turns up at least one thing that is switched off by accident.
→ AI search optimization → Answer engine optimization services → Get a free audit
All posts →
SEO & Traffic 10 min CPA Marketing: The Clients Worth Having Are the Most Expensive Clicks in Accounting Cost per click inside accounting runs from $6.65 to $258.68 — a 39x spread in one profession. The divide is consumer versus business, not tax season. Read article → SEO & Traffic 9 min Veterinary Marketing: Why Denver Clinics Pay Triple for Every Click All 16 veterinary “near me” searches cost more in Denver than nationally — median +138%. What an independent clinic does instead of bidding. Read article → SEO & Traffic 11 min The Denver Click Premium: What a Customer Actually Costs in 40 Local Industries Original research: we priced 40 local service industries nationally and in Denver. The same click costs more here in 32 of 37. Read article →
Ready to apply this?
Get a free AI marketing audit — we'll analyze your current setup and show you exactly where AI can move the needle fastest.
Get a Free Audit Explore Services
Custom marketing agents. Human oversight. Systems that keep working. Built in Denver since 2018, run everywhere.
COMPANYHomeAgentsBuildAiden OSPricingAbout
EXPLORECustomersHow It WorksWho It’s ForAEO ServicesBlogFAQ