Why does ChatGPT never quote our blog, and how do we fix it?

Your firm has been blogging for years. There are two hundred posts on the site; some rank and a few bring in enquiries. Then someone asks ChatGPT the question one of those posts was written to answer, and the assistant quotes a trade body, a national firm and a software vendor. Try Gemini, Perplexity and Google's AI Overviews, and the result is the same.

The usual explanation is that AI "judges content differently from Google". That is only half true. ChatGPT search, Perplexity, Copilot and Google's AI features start by running web searches, so they mostly quote pages they can find. In an AirOps study reported by Search Engine Land in April 2026, pages in the top search position were cited by ChatGPT 58.4% of the time, against 14.2% for pages in position ten.

So an ignored post fails at one of three gates: the assistant cannot reach it, it never turns up in the searches the assistant runs, or it turns up and gives the assistant nothing worth quoting. Check them in order, because a rewrite fixes nothing if the first two are closed.

#Why does ChatGPT cite a competitor's blog but not ours?

Their post passed all three gates; yours stopped at one. Work through these checks in order and stop at the first that fails.

1. Can AI crawlers reach the post? Check robots.txt and your CDN. If not, fix access before anything else.
2. Is the text in the page's HTML? View the page source and search for your first paragraph. If it is missing, the text is rendered by JavaScript, which the main AI crawlers do not run.
3. Is the post indexed in Google and Bing? Search each for the post's exact title. If it is missing from Bing, submit your sitemap in Bing Webmaster Tools.
4. Does it rank for the question and its follow-ups? Search the buyer's question and two or three related ones. If you are nowhere, it is a ranking problem as much as a writing one.
5. Is the question one a blog post can answer? If buyers ask "who are the best firms for X", the answer comes from lists, reviews and directories.
6. Only now: is it a content problem? If the post passes all five checks and is still ignored, the four content reasons below apply.

#Is ChatGPT blocked from my website, and how do I check?

Open yoursite.co.uk/robots.txt and look for the AI crawlers by name. OpenAI says a site must allow OAI-SearchBot, and its host or CDN must allow OpenAI's published IP addresses, to be eligible for ChatGPT search. The names that matter:

Then check your CDN. Cloudflare changed its default to block AI crawlers in July 2025 and updated its defaults for new domains again on 15 September 2026, so look at the AI bot settings in your dashboard even if nobody on your side chose to block anything.

Finally, check JavaScript. Vercel and MERJ found in December 2024 that GPTBot, ClaudeBot and PerplexityBot fetch JavaScript files but do not run them, so a blog whose text loads by script is blank to them. Googlebot does render JavaScript, which helps Gemini and Google's AI features. Our guide to how AI crawlers read your website covers each check in detail.

#Does ChatGPT use Bing or Google to find the pages it cites?

It uses partners, and Microsoft is one of them. OpenAI's help page says ChatGPT search "sometimes partners with other search providers" and rewrites your question into targeted queries it sends them, and it lists Microsoft among those partners. In a small Seer Interactive test in February 2025, 87% of ChatGPT search citations matched Bing's top results. Copilot draws on Bing; Gemini, AI Overviews and AI Mode draw on Google's index.

So a post missing from Bing can be invisible to ChatGPT and Copilot. Check Bing Webmaster Tools as carefully as Search Console.

#Why can a post rank on Google and still not be quoted?

Usually because it ranks for the main query but not for the related searches the assistant actually runs, or because it ranks and has nothing quotable in it. Ranking and citation are closely linked, as the AirOps figures show, but they are not the same. Ahrefs found in March 2026 that only 38% of AI Overview citations came from pages in Google's top ten, down from 76% in July 2025, because Google now runs several related searches for one question. A post that answers only the headline question misses the follow-ups.

The fix is to cover the follow-up questions a buyer asks, in their words. If the post ranks for them and is still passed over, it is a content problem.

#Is the question one a blog post can answer?

Often it is not. Most blog posts answer informational questions, such as "how does X work?" or "what does X cost?". Buyer prompts that name suppliers, such as "best commercial property solicitors in Leeds" or "which firm should we use for X?", are answered from list articles, review sites, directories and press coverage. In G2's April 2026 survey of 1,076 software buyers, 45% said citations of review sites were the most confidence-inspiring signal in an AI answer.

Comparison pages are the bridge. For Helm Club, "We built the structured data AI engines read, then a comparison page for every question a founder asks when weighing up peer groups: Helm against YPO, Vistage and the rest." And "Every comparison page we have published ranks first for its question on Google and the AI engines. The pages not yet built are the ones where a competitor still writes the answer." For shortlist questions, see why ChatGPT sounds unsure about your firm.

#If the post is found, why is it still not quoted?

In our experience, for one of four reasons. To prioritise across many posts, use the scoring rubric in our guide to auditing existing content, and for sites with hundreds of near-duplicate posts, read why ChatGPT ignores sites with hundreds of pages.

#Reason 1: the post says nothing specific

The most common failure by a distance. The post covers the topic without committing to anything: "Data governance matters. Firms should implement governance to protect data quality." All true, all useless, and it could sit on any competitor's site unchanged.

Compare: "In the governance projects we have run, the ones that stalled almost always created policies before changing who was allowed to use which data." That is a mechanism, a pattern and a judgement an assistant can use and has to attribute.

How to tell. Mark every sentence a competitor could have written. If most are marked, the post is generic.

How to fix. Add figures from your own work, named patterns, the reasons things happen and where the advice does not apply. In a Princeton and IIT Delhi study presented at KDD 2024, adding statistics raised visibility in AI-generated answers by about 34%; it was tested on a research engine, so treat it as directional.

#Reason 2: nothing shows who wrote it or why they know

No author, or "by the team". No example that could only come from experience. No sources, no limits. Google's guide to its generative AI features, updated in July 2026, says content with "unique expert or experienced takes" counts for more than commodity content, and an anonymous post rarely shows either.

How to tell. Is there a named author with a role? An example only a practitioner would know? A cited source? A paragraph on where the advice fails? Three noes is a weak post.

How to fix. Assign each subject to one named person and put a proper byline on every post: name, role and a sentence of experience, linked to a profile page. Add two or three examples from practice, anonymised if needed, cite borrowed facts and say where the advice stops working. What makes ChatGPT treat content as expert rather than marketing goes further.

#Reason 3: the structure makes it hard to use

A post that wanders, with no clear headings and the answer buried at the end, is hard to lift from. A 2026 analysis of 1.2 million ChatGPT answers found 44% of citations came from the first 30% of a page, so a post that saves its point for the conclusion wastes its best sentence.

How to tell. Can you outline the post from its headings in two minutes? Is the answer in the first two hundred words?

How to fix. Lead with the answer, turn each heading into the question it answers and keep sections short. What an article that ChatGPT actually quotes looks like sets out the full shape and the evidence for it.

#Reason 4: it has no point of view

A post that agrees with everyone gives an assistant no reason to choose it over the fifty others that say the same. A point of view is something noticed from the work: that the usual advice fails in a common situation, or that one step matters far more than the others.

How to tell. Would a competitor read it and say "that is what I would have written"? Does it disagree with received wisdom anywhere?

How to fix. Ask the practitioners what they believe the industry gets wrong, and what they have seen that supports it, then rebuild the post around that. This is the hardest fix, because it needs thinking as well as editing.

#How do we rewrite a post?

Do this for the twenty posts that matter most: those on your core services, those that still get traffic and those that will stay relevant. In our experience the whole process takes ten to twenty hours a post, which is why you choose twenty rather than two hundred.

#Step 1: Diagnose it

Run the six checks above and write down which gate the post fails. About an hour.

#Step 2: Gather what it is missing

Talk to the practitioner, pull figures from your own records and find the sources worth citing. This is the step most people skip, and where the value is. Three to eight hours.

#Step 3: Outline the new version

The buyer's question as the title, the answer in the first paragraph, the follow-up questions as headings. About an hour.

#Step 4: Rewrite it

Write it fresh rather than patching the old text. Keep the specifics, cut the filler, add the reasons and the exceptions. Four to ten hours.

#Step 5: Add the structure and the author

Key points near the top, the buyer's follow-ups at the end, links to related pieces, a proper byline and a last-reviewed date. One to two hours.

#How do I see whether ChatGPT has ever cited our site?

Partly from analytics, fully only by asking. GA4 added a default "AI Assistant" channel on 13 May 2026 covering visits from ChatGPT, Gemini, DeepSeek, Copilot and Grok, though not Claude or Perplexity, and clicks from AI Overviews and AI Mode count as Organic Search. Search Console's generative AI reports show your impressions in AI Overviews and AI Mode. Your server logs show ChatGPT-User visits, which happen when ChatGPT fetches a page to answer someone.

None of that shows a citation nobody clicked. For that, keep a fixed list of 20 to 50 buyer questions, ask each several times a month in a clean session, and count a win only when it repeats. How to find out whether ChatGPT is recommending your firm sets out the routine.

#Should we stop blogging and put the effort into other channels?

No, but blog less and spread the effort. Posts that answer real buyer questions still feed the informational answers. The shortlist answers, though, lean on what others say about you: Muck Rack found in May 2026 that earned media made up 84% of the links AI tools cited. A sensible split is fewer, better posts, plus reviews, trade coverage and LinkedIn articles by named experts. Researched outbound, with every message written and approved by a person, reaches buyers before any assistant is asked, and works alongside all of it.

#What to do this month

1. Check robots.txt, your Cloudflare or CDN settings and the page source of three posts, and fix any access problem first.
2. Check that your twenty most important posts are indexed in Bing as well as Google.
3. For each of the twenty, search the buyer's question and two follow-ups, and note where the post ranks.
4. Sort the twenty by gate: access, ranking, wrong question type or content.
5. Book practitioner time for the top five content rewrites and start the first this week.
6. Build your fixed list of buyer questions and record who is named today, as the baseline.

#Frequently asked questions

#Should we delete the posts we are not rewriting?

Not if they bring traffic or help a visitor. Keep what earns its place, and merge or retire posts that nobody visits and that repeat each other. The priority is the twenty that matter, not the tidiness of the archive.

#Can we fix one thing without a full rewrite?

Sometimes. Access, indexing, bylines, headings and sources can all be fixed without touching the argument. Specificity and point of view cannot, because they change what the post says. If the diagnosis points at those, plan for a rewrite.

#What if every post we have is generic?

That is common, and it usually means the expertise never made it onto the page. Ask the partners what they believe that competitors would argue with, and what supports it. That conversation produces the first five rewrites.

#Can an agency do the rewrites?

An agency can do the structure, the sourcing and the writing, provided it has access to your practitioners. The arrangement that works is a writer interviewing an expert for forty minutes per piece, then drafting. The one that fails is a writer given a keyword and no expert.

#How many posts a month should we publish once the rewrites are done?

Fewer than you think. One thorough piece a month by a named expert, answering a real buyer question, does more for a reader and for being quoted than four thin ones. How often we need to publish for AI assistants to notice us covers the evidence on frequency.

If you would like an outside view of which gate your posts are failing, book a free Growth Review: 30 minutes with David Adams and three fixes in writing. Turning a firm's expertise into content that gets it quoted is the work of Reputation That Wins.

#Sources