We asked Perplexity two questions our articles answer
We asked Perplexity two questions ixsoftum articles answer, in September and again in October. It never cited us. Here's what it cited, and the likely reasons.
- Across two runs, Perplexity didn't cite ixsoftum for either question, even though both questions match ixsoftum articles.
- For a product question, it cited almost only the vendor: in October, all 10 sources for the Claude Code auto mode question were Anthropic pages.
- One of our two articles has never been indexed by Google. A page a search engine hasn't indexed is unlikely to be cited by a search-backed AI tool, whatever its quality.
- Two prompts on one site can't show what wins citations in general. Run the same check on your own pages, and check indexing first.
AI search tools such as Perplexity show which sources they used, which makes it possible to check whether your content gets cited. We ran that check on ixsoftum with two questions our articles answer directly, first in September and again in October 2026. Perplexity cited ixsoftum in neither run.
This piece reports what we found, what it can and can't tell you, and the cause we missed the first time.
What we tested
We picked two questions that match ixsoftum articles:
- "what does claude code auto mode actually do", which matches What Claude Code auto mode actually does;
- "is vibe coding bad", which matches Is vibe coding bad?.
We used Perplexity because it lists its sources next to each answer. The first run was on September 7, 2026, and we didn't keep the URLs. The second was on October 23, 2026, with default settings, and we kept the full source lists.
What Perplexity cited
For the auto mode question, Perplexity cited 15 sources in September. Anthropic's own pages appeared most often, alongside TechCrunch, a few tech news sites and several independent blogs. In October, all 10 sources were Anthropic's: Claude Code documentation, including three translations of the same page, Anthropic's engineering post on how auto mode was built, and articles on claude.com.
For the vibe coding question, the sources were more varied both times:
- September: Hostinger, The Conversation, SoftwareMind, Wikipedia, Codecademy, an independent blog, Hacker News and Reddit.
- October: the UK National Cyber Security Centre's guidance, Tanium, IBM, The Conversation, InfoWorld, Lifehacker, Codecademy, AOL and a Hacker News thread.
The October sources for the auto mode question were:
- Configure auto mode, plus its Russian, French and Spanish versions;
- Choose a permission mode;
- Auto mode for Claude Code;
- Auto mode is now the default;
- Running auto mode in production;
- How we built Claude Code auto mode, in two places on anthropic.com.
Answers vary between runs and users, so treat these as two snapshots, not a stable ranking.
The cause we missed in September
The September version of this article put the result down to authority and age. It missed a more basic explanation: indexing.
On October 23, Google Search Console's URL Inspection reported the auto mode article as "URL is unknown to Google". It has never been indexed, despite being submitted. A search-backed AI tool retrieves pages from a search index, so a page that isn't in the index is unlikely to be cited however accurate it is. The vibe coding article is indexed (last crawled October 8) and still wasn't cited, so indexing isn't the whole story. It is the first thing to rule out.
What two prompts can and can't show
The vendor pattern on the auto mode question is clear: for a question about a product, the product's own documentation took every citation. That fits how these tools are built to prefer primary sources, and it's the most useful result here. If you make a product, your documentation is the page most likely to be cited for questions about it, so keep it complete and current.
The vibe coding results show a mix of government guidance, vendors, mainstream tech press, reference sites and community discussion. Two runs of one question can't say which of those signals matters most, or what a new site would need to be cited.
The September version also drew wider conclusions than this sample supports: that citation is won by primary authority, institutional trust or community discussion, and that no optimization would have changed the result. We've removed them.
How to check your own AI visibility
The same check works for any site, and it costs nothing:
- Check indexing first. Use Search Console's URL Inspection on the pages you expect to be cited. If a page isn't indexed, fix that before anything else.
- Pick questions your pages answer directly, and ask them in a tool that shows its sources, such as Perplexity.
- Record the full source list with the date, and repeat the check over a few weeks. Single runs vary.
- Look at who wins for each question. If it's the vendor's own documentation, that's hard to displace. If it's community threads or press coverage, earning that coverage may matter more than editing your page.
Correction, October 23, 2026: an earlier version of this article said "I ran" the test under a named byline; it was run by the ixsoftum team. It also drew general conclusions from two prompts and missed that one tested article wasn't indexed by Google. We re-ran the test, published the sources, and narrowed the conclusions.