Plugin Visibility Index · Run 06
Zero citations: what a month of publishing did not do
On 10 September, in our baseline run, we asked AI search engines 24 real questions about AI visibility and WordPress plugins, and softrecs.com was cited in none of them. We set a target before doing anything: cited in 2 to 4 of 24 by 10 October, and if the month ended at zero, change the approach instead of repeating it. This is the 10 October measurement. The answer is still zero.
- 0 of 72 answers cited softrecs.com. None named it in the text either.
- ChatGPT read 385 sources and showed 47 (12%). The median answer retrieved 16.5 and displayed 1.
- Google AI Mode leaned on platforms, not websites: YouTube appeared in 20 of 24 answers, Reddit in 19, LinkedIn in 13.
- The engines barely agree. Median source overlap per prompt: ChatGPT vs Google 2.9%, ChatGPT vs Perplexity 3.9%, Google vs Perplexity 10.6%.
What we ran
24 prompts in five clusters: symptom questions (6), WordPress (3), data and sources (8), myths (4) and definitions (3). 22 are Google autocomplete suggestions pulled on 7 October. Two WordPress prompts are written by us, because Google suggests no WordPress-specific phrasing for this topic. None of the 24 repeats an earlier run’s wording: Run 04 showed that an identical prompt can return an identical, cached source list, so a repeat would measure the cache instead of the engine.
Each prompt was asked once per engine. ChatGPT: a temporary, unpersonalized chat with web search on, with the full retrieval list read from the conversation record. Google AI Mode: US English, with the full source panel opened. Perplexity: default mode, sources read from the Links tab. As a positive control, the branded query “softrecs plugin visibility index” on Perplexity does cite softrecs.com, so the method can see the site when it is there.
One note on timing: Perplexity’s free tier hit its search limit after 11 prompts. The next 10 Perplexity answers were collected later on 10 October and the last 3 on 13 October. ChatGPT and Google AI Mode ran all 24 on 10 October.
Where the sources came from
| Engine | Answers | Median sources shown | Most frequent source |
|---|---|---|---|
| ChatGPT | 24 | 1 (16.5 read) | help.openai.com, 10 of 24 |
| Google AI Mode | 24 | 12.5 | youtube.com, 20 of 24 |
| Perplexity | 24 | 9 | blog.hubspot.com and semrush.com, 9 of 24 each |
softrecs.com was a source in none of them. The three engines form three different source pools. Across all answers Google used 180 distinct domains and ChatGPT 281; only 29 domains appear in both. Perplexity used 119, and just 16 appear on all three engines. Google’s pool is dominated by user platforms: YouTube, Reddit, LinkedIn and Quora together. ChatGPT used those same platforms almost never (Reddit, YouTube and LinkedIn once each in 24 answers) and leaned on vendor documentation instead: on questions about ChatGPT, it retrieved openai.com pages in 10 answers.
The part that matters most for a small site: ChatGPT’s retrieval lists were full of small, recent GEO sites, several of them clearly younger than this one. Being new and small is not what kept softrecs.com out. ChatGPT search is widely reported to rely on Bing’s index, and Bing Webmaster Tools still lists the softrecs.com homepage, /index/ and /leaderboard/ as discovered but not crawled. A page Bing has not crawled is a page ChatGPT cannot retrieve.
What changes now
We said that zero would mean the approach was wrong, so it changes here. Publishing more guides on our own domain was the September plan; this data says that alone does not reach either engine.
- Bing first. Get the core pages crawled in Bing, because that is the retrieval path ChatGPT uses.
- Go where Google AI Mode reads. The original data in this index goes onto YouTube, Reddit and LinkedIn as well as this site, since those three were in most Google answers.
- Keep measuring the same way. The next run uses new phrasings again, and this page links to it when it exists.
The October leaderboard run
The same day we re-ran the AI Plugin Leaderboard with 15 new phrasings across booking, caching, backup, LMS and membership plugins, on ChatGPT and Google AI Mode (Perplexity is pending for the same reason as above). The September patterns held. UpdraftPlus was named first in all 6 backup answers. FS Code domains (fs-poster.com, fs-code.com, booknetic.com) were a source in 7 of 30 answers, in categories where FS Code sells nothing. gigapress.net appeared in 7 of 15 ChatGPT retrieval lists and in no Google answer. All-in-One WP Migration, with over five million installs, was named in 2 of 6 backup answers.
Data
Every answer is one row in the CSV: prompt, cluster, engine, whether softrecs.com was cited or named, sources shown, sources retrieved (ChatGPT) and the domains. Download the Run 06 CSV · Download the October leaderboard CSV. Licensed CC BY 4.0.
What the engines say about themselves: How Perplexity works, Google’s AI features and your website (AI Overviews and AI Mode), and OpenAI’s Searching the web with ChatGPT. Those pages describe what each engine intends; the figures here record what it actually did.
How this was made: figures come from our own runs in the engines’ public interfaces on 10 October 2026. Drafted with an AI assistant (Claude) and reviewed by Alauddin Chowdhury. Method and AI use.
Earlier in this index: Run 05, ChatGPT as a third engine.