Earning the Shortlist · Published October 27, 2026
My predictions for Earning the Shortlist
I'm writing these down before the first formal round. If I only report results after I see them, every result looks like it was the plan. So here's what I expect, how sure I am, and how I'll score myself. This page doesn't change after October 27, 2026; corrections go underneath, dated.
Disclosure, October 27, 2026An exploratory API pilot ran on October 20, before these predictions were public. It asked the ten questions once each through the ChatGPT, Perplexity and Claude APIs (30 answers). That's a different method from the consumer apps I'm tracking from now on, so I won't score these predictions against it, and I'm publishing them with today's date, not an earlier one.
What I'm basing this on
- My site gets seen by the wrong people. Search Console: my AI visibility tools list earned 6,274 impressions and 2 clicks between July 16 and October 16. Tool shoppers, not buyers.
- I already have pages for two of the problem prompts. "How to get your B2B SaaS cited by ChatGPT" (question 7) and "SEO vs AEO for B2B SaaS" (question 8). Question 9 is partly covered by my content-system posts.
- I have no pages for the who-to-hire or comparison prompts. Nothing aimed at questions 1-6.
- I'm a small, new brand. Who-to-hire answers lean on third-party mentions: lists, reviews, Reddit, directories. I have very few of those.
Where I think I stand today
Scored at the end of the two-week baseline.
| # | Prediction | Confidence | Result |
|---|---|---|---|
| B1 | My own name (question 10): AI describes SuperMarketers accurately (what I do, who it's for) in at least 6 of 9 answers. | 65% | Not scored yet |
| B2 | Who to hire (questions 1-4): SuperMarketers is recommended in 2 or fewer of 36 answers. | 85% | Not scored yet |
| B3 | Comparisons (questions 5-6): SuperMarketers is mentioned in 1 or fewer of 18 answers. | 80% | Not scored yet |
| B4 | Problem prompts (questions 7-9): a supermarketers.ai page is cited in at least 2 of 27 answers. | 55% | Not scored yet |
| B5 | Surfaces: Perplexity cites my pages more often than ChatGPT or Gemini. | 60% | Not scored yet |
| B6 | Who gets recommended instead: most agencies named in the who-to-hire answers are established SEO or content agencies, not AI-native implementation shops. | 65% | Not scored yet |
| B7 | Consistency: at least half of all prompt-surface pairs give a different set of recommended companies across their 3 runs. | 70% | Not scored yet |
Week 10, after eight weeks of focused work
| # | Prediction | Confidence | Result |
|---|---|---|---|
| W1 | A real citation win: at least one problem prompt (questions 7-9) cites a supermarketers.ai page in 2 or 3 of 3 runs, on at least one surface, for 3 weeks in a row. | 50% | Not scored yet |
| W2 | A who-to-hire breakthrough: SuperMarketers is recommended for at least one who-to-hire prompt (questions 1-4), on any surface, in 2 consecutive weeks. | 20% | Not scored yet |
| W3 | Branded fixed: question 10 is accurate in at least 8 of 9 answers. | 70% | Not scored yet |
| W4 | Listicles (Experiment 04): the buyer-facing list gets more Google impressions than its matched non-list page over 8 weeks. | 60% | Not scored yet |
| W5 | Listicles, clicks: that list still earns 10 or fewer clicks over the same 8 weeks. | 70% | Not scored yet |
| W6 | Search Console overall: clicks to www.supermarketers.ai stay under 20 per 28 days at week 10. | 75% | Not scored yet |
| W7 | The stop rule: I keep paying for tracking after week 10 (W1 or W2 came true). | 55% | Not scored yet |
How I'll score myself
- Each prediction is true or false, judged from the published answers (the receipts on the Lab), not from my memory.
- I'll publish the hit rate for each confidence level: of my 70% calls, how many came true?
- No edits after publishing. If I think a prediction was badly worded, I'll say so underneath, dated, and still score it as written.
What you can take from this
- Write predictions before you measure. It's the cheapest protection against fooling yourself with your own dashboard.
- Separate your prompts by how hard they are to win. Branded is days. Problem prompts are weeks. Who-to-hire is months, because it depends on other people mentioning you.
- Put a confidence on every guess. "I think we'll show up" isn't a prediction. "55% that we're cited in 2 of 27 answers" is.