We publish 20 blog posts a month and none get cited by ChatGPT or Perplexity: what are we doing wrong?
Twenty posts a month with no AI citations is usually a crawl, prompt-match or originality problem, not a volume problem. Here is how we diagnose it for clients.
This is one of the most common conversations we have. A team publishes twenty posts a month, sees nothing from ChatGPT or Perplexity, and concludes they need to publish more or write longer. Usually the opposite is true. Assistants do not reward cadence. They retrieve pages that match a specific question, can be fetched cleanly, and say something worth quoting. One of those is failing, and it is a diagnosis job, not a writing job.
We do not cite studies about what AI search favors in these conversations. We have no source we could stand behind, and the answer is on the client's own site anyway. Here is the order we work in.
1. Can the bots read the posts?
We always start here, because it is a yes or no question and it can invalidate everything else. We look at robots.txt for the bots that matter (OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User and others), then at CDN, firewall and bot-protection rules that may refuse them regardless. We check whether the blog renders without JavaScript and whether the bots receive errors, redirects or timeouts.
In Promptwatch, Agent Analytics gives IP-verified crawler logs, so spoofed user agents are filtered out and we see real bot traffic. The crawl-to-citation path shows a citation rate per page, and error tracking gives failure explanations. Crawler logs start on the Professional plan, which we flag in scoping.
Promptwatch's TreeHouse case study is a good illustration of order of operations. TreeHouse, which runs Pararius and Huurwoningen.nl, reports that AI search became about 25% of its organic conversions, 99% of it from ChatGPT, after fixing crawler access and automating weekly reporting with the Promptwatch MCP. That is the customer's own result, and we attribute it to them. The lesson we take is to fix access before editing anything.
2. Do the posts match real prompts?
A calendar built from keyword tools can fill a month with posts no one asks an assistant about. Assistants split a question into narrower searches, called query fan-outs, and retrieve pages that match those. We pull the buyer prompts, look at the per-model fan-outs in Promptwatch, and compare them with the client's post titles. The mismatch is often clear within an hour.
Content gap analysis then scores coverage against the tracked prompts. It tells us which topics have no good page, which is a better guide to what to write than a keyword spreadsheet.
3. Do the posts say anything new?
Twenty posts a month can easily be twenty restatements of what the top results already say. An assistant has no reason to quote a page that adds nothing. So we ask clients what they know that others do not: customer questions, real process details, product behavior, pricing logic, small data sets they can stand behind. Those become the spine of a post. We also push for a direct answer in the first paragraph, named sources and dated facts.
4. Is anyone else naming the brand?
Assistants lean on third-party pages, and a blog speaking alone is easy to ignore. The offsite mentions report in Promptwatch shows where a brand is named without a link. Citation analytics show which domains get cited for the client's prompts, including Reddit, YouTube and LinkedIn reports. If competitors appear on sources the client is absent from, part of the plan is outreach and presence there, not another blog post.
5. Are we measuring the right thing?
Sometimes the posts are cited and nobody sees it, because the relevant prompts are not tracked. Sometimes the posts are fetched and read but never cited. Those need different fixes. Page Tracker lets us follow specific URLs, ours and a competitor's, so we can see which page keeps winning a prompt and read it.
What we change in the content program
Our usual recommendation is to cut the calendar and deepen each piece. Five strong, original pages aimed at tracked gaps beat twenty generic ones, and they cost less. For clients who do want scale, Promptwatch has a Content Agent on Business and Enterprise that plans, writes and publishes against visibility gaps on a schedule, with a review inbox so an editor approves each piece. Crisp, a customer, reports that AI-referred traffic converted at twice the rate of traditional channels and that it scaled to 5 to 10 articles a day. That is Crisp's published case. We do not present it as a typical result, and we still read every draft ourselves.
The 30-day version
- Week 1: access audit and visit tracking.
- Week 2: a prompt set of 30 to 50 questions, mapped against existing posts.
- Week 3: the five most relevant gaps, written with original material.
- Week 4: re-measure crawler fetches, citations and AI-referred visits for those pages.
What we tell clients honestly
Nothing guarantees a citation. Engines change, answers vary between askings, and competitors publish as well. What the process gives you is a reason for the silence, which is more than most teams have after a year of posting. If the reason is access, you fix it in days. If it is originality, it takes longer, and that is a decision about how the company wants to use its expertise.