Guide 03 · Diagnostic
Why your plugin isn’t showing up in ChatGPT
There are three different reasons, they look identical from the outside, and they need opposite fixes. You can tell them apart in about fifteen minutes. Start at the top — most people skip the first one and it is the one that is usually broken.
Is this actually common?
Yes. Across ten WordPress plugin categories I tracked 52 plugins and asked both Perplexity and Google AI Mode the ordinary buying question for each category. Ten of the 52 — 19% — were named in no open-ended answer on either engine. Not ranked low. Not mentioned.
The one worth staring at is All-in-One WP Migration. It has over five million active installs. Across six Perplexity answers and one Google answer it was named exactly once — in the answer to “UpdraftPlus vs All-in-One WP Migration”, where the question supplied the name.
So this is not a small-plugin problem, and it is not a quality problem. Five million installs buys you nothing here. Whatever is happening is mechanical, and mechanical things can be diagnosed.
The three reasons, in the order to check them
| Reason | What it means | What fixes it |
|---|---|---|
| 1. Unreachable | The crawler asks your server and does not get your content | A server change, usually one afternoon |
| 2. Unknown | The engine has nothing about your plugin to draw on | Content and mentions, months |
| 3. Unretrieved | The engine knows you, and does not pull you up when nobody names you | Getting into the pages it reads instead |
Almost all published advice addresses number two. Number one is more often broken than anyone expects and is far cheaper to fix. Number three is the one most established plugins actually have, and writing more about your own features does not touch it.
Reason 1: can the crawlers reach you at all?
Check this first, because it is binary. If a retrieval crawler cannot fetch your page, nothing else on this page matters — no content strategy operates on a page that never arrives.
From a terminal:
curl -A "OAI-SearchBot" -I https://yoursite.com/
curl -A "PerplexityBot" -I https://yoursite.com/Anything other than a 200 at the end of the redirect chain is your answer, and you can stop reading. The four things that break this, in the order I find them:
Without a terminal, the free retrievability check runs all four of these for you from a wordpress.org slug.
| Failure | How it looks to you | How it looks to a crawler |
|---|---|---|
| Expired TLS certificate | Browser warning you clicked past months ago | Fetch fails. Total. |
| WAF or CDN bot rule | Site is fast and fine | 403, even though robots.txt allows it |
| Client-rendered content | Page looks perfect | An empty shell |
| robots.txt blocking retrieval bots | You blocked “AI scrapers” once | Excluded from answers, not from training |
The certificate case is not hypothetical. This site was invisible to every retrieval crawler for 152 days on a certificate that had lapsed, while the content sat there being fine. Nothing in a content audit surfaces that.
Training crawlers and retrieval crawlers are different bots with different names. GPTBot and ClaudeBot collect training data — blocking them is a legitimate business decision. OAI-SearchBot, PerplexityBot, Claude-SearchBot and Bingbot fetch pages so an answer can cite them. Blocking those does not protect anything; it removes you from the answers. Plenty of sites blocked the whole group in one line and did not realise they had opted out of being cited.
Two more that are easy to miss. A robots.txt allow is not the same as a server 200 — bot protection at the CDN sits in front of robots.txt and will refuse the request anyway, so the live check above is the one that counts. And Googlebot renders JavaScript while retrieval crawlers generally do not, so a site can rank perfectly well in classic search and be invisible to the engines that answer questions. Fetch a real page and read the raw HTML. If your content is not in it, the crawler does not see it either.
What is not on this list is llms.txt. It is the most commonly recommended fix for this exact symptom and there is no evidence any engine requests it — the numbers are here.
Reason 2: does the engine know your plugin at all?
This is the fastest test on the page. Ask an engine about your plugin by name: “what is [your plugin] and what does it do?”
Read the answer for accuracy, not for flattery. If it describes your features correctly, names your pricing tiers roughly right, and knows what it competes with, the engine is not ignorant of you. That is a good result and it changes what you should do next — it rules out an entire category of work.
Both produce “my plugin never comes up”. They need opposite fixes. If the by-name answer is good, you do not have a content problem, and no amount of new writing about your own features will solve it. If the by-name answer is wrong or empty, you do — and that is a different, slower job.
Asked directly about All-in-One WP Migration, Perplexity produced a detailed and accurate comparison. The engine knew the product perfectly well. It simply never reached for it when the buyer had not already thought of the name — which is every discovery question that matters commercially.
One correction worth running while you are here: check whether the answer describes a version, a price or an owner that is out of date. Engines hold stale product facts for a long time, and they hold different stale facts from each other. Perplexity named a sunset membership plugin in all three of its membership answers, twice in third position, once calling it the best lightweight option. Google AI Mode did not name it at all. If your product was renamed, acquired or repriced, the correction has to be checked on each engine separately rather than once.
Reason 3: you are known, and still not retrieved
This is the common case for an established plugin, and it is the hardest one, so it is worth being precise about what is actually happening.
When someone asks “best WordPress backup plugins”, the engine does not consult a ranked list of plugins. It retrieves a set of pages and writes an answer from them. If you are not in those pages, you are not in the answer — however good your product is and however much the model knows about it in the abstract.
So the question stops being “how do I rank” and becomes “which pages get read for my category, and am I named in them”. That is answerable. Open the source list on the answer and look at what is actually there.
What those source lists look like
| Measured across ten categories | Perplexity | Google AI Mode |
|---|---|---|
| Sources owned by a plugin vendor | 41% | 47% |
| Categories citing youtube.com | 0 of 10 | 8 of 10 |
| Source overlap between the two engines | 15% (median) | |
Three things follow from that table, and they are the practical core of this page.
Roughly half of every answer comes from vendor-owned domains. Not independent review sites — companies selling WordPress products, writing about categories including ones they do not sell in. Your own writing is eligible on both engines. The entry is open.
YouTube is a Google-only channel, and a large one. Google AI Mode cited youtube.com in eight of ten categories; Perplexity in none. If you are missing from Google’s answers specifically, and you have no video, that is the most concrete gap in this data. No amount of writing substitutes for it there.
Being missing from one engine is a real, separate problem. The two engines share a median 15% of their sources, and rewording a question inside one engine moves only about a fifth of them. The gap between engines is roughly four times larger than the noise within one. So if you are cited on Perplexity and absent from Google, that is not bad luck and it will not average out — it needs its own fix.
What does work, then?
One pattern in this data demonstrably travels across both engines. A single publisher network — a booking-software company — was cited by Perplexity in seven of ten categories and by Google in eight, and in six or seven of those it sells nothing at all. In a corpus where the two engines agree on 15% of sources, that network is in the small set they agree on.
What it is doing is not a backlink strategy and not schema. It is a large volume of long, structured, comparison-shaped articles about categories adjacent to its own product. That is expensive in time and cheap in money, which is the right shape for a plugin company.
Two cheaper things worth knowing while you plan that. Brand mentions correlate with AI visibility at 0.66 and backlinks at 0.19 — an unlinked mention of your plugin in a listicle or a forum thread still puts your name in the text the model reads, and a link is a bonus rather than the payload. And listicles account for 21.88% of citations against 2.2% for comparison pages, so the instinct to generate a /vs/ page for every competitor aims most of the effort at a fiftieth of the opportunity.
The fifteen-minute version
| Step | Do this | If it fails |
|---|---|---|
| 1 | curl -A "OAI-SearchBot" -I your site | Stop. Fix the server. Cheapest win available. |
| 2 | Fetch a page, search the raw HTML for your own text | Server-render it. Binary problem. |
| 3 | Ask an engine about your plugin by name | Answer wrong or empty — you have a content problem |
| 4 | Ask the category question, open the full source list | You are absent from the pages being read |
| 5 | Repeat step 4 on the other engine | Engine-specific gap. Fix separately. |
One caution on measuring this repeatedly. Perplexity returns an identical source list for an identical query string — in the categories I tested, ten of ten and twelve of twelve domains, hours apart. Adding a single question mark changed the sources every time. So if you plan to re-check monthly, vary the phrasing, or you will be reading a cache and calling it a trend.
What this page cannot tell you
Ten categories, 52 plugins, two engines, September 2026. The percentages are from a small base and the engines change without notice. The diagnostic order holds under every reading of this data; the individual figures should be read as illustrations of a shape.
The correlations quoted here are observational. They cannot separate cause from a common cause, and brands that get mentioned also tend to get linked and to have video. What the ordering tells you is where the cheap wins are not, which is a weaker claim than a causal one and the only one the evidence supports.
It also cannot tell you that fixing reason one will get you cited. It will not, on its own. Retrievability is necessary and nowhere near sufficient — this site has been retrievable and correct for weeks and is cited in none of the twenty-four prompts it tracks. I publish that count either way.
Not sure which of the three you have?
Send me your wordpress.org slug. I will run the category question on both engines, pull the full source lists, and tell you which of the three reasons you are actually dealing with — and what the first fix is. Free, no call, usually within a day.