AI Answer Visibility Index, 2026-Q4: what AI answer engines cite when founders ask about marketing

AAJ asked Claude, Perplexity and OpenAI's GPT-4.1 the same 40 founder marketing questions: how often they search, whom they cite, and whether AAJ shows up. By Saroj Jha, AAJ, 28 September 2026.

Download the data (CSV) — Measure your own brand with the AI Visibility Significance Test

Key findings

• 100% vs 5% of answers cite any source. Perplexity cited sources in 40 of 40 answers, Claude in 10 of 40 and GPT-4.1 in 2 of 39.
• 178 different domains in 241 citations. No domain earned more than 6 citations (2.5%). The ten most-cited domains share 17%.
• 0 citations to the big research firms. Gartner, Forrester, McKinsey, HBR and the other research sources on AAJ's list were never cited.
• 0 of 119 answers named or cited AAJ. This is the baseline. With 119 answers, the true rate is below 3% at 95% confidence.

The short answer

Whether an AI answer can cite you depends first on whether the engine searches at all. Perplexity cited sources in every answer. Claude cited sources in 10 of 40 and OpenAI's GPT-4.1 in 2 of 39: for most of these everyday marketing questions, they answered from what they already knew.

When the engines did cite, they spread it thin: 241 citations went to 178 different domains. No publisher dominates founder marketing questions, which is good news for a small firm with something specific to say.

The sources that did get cited are mostly companies writing about their own category. 10 of the 14 domains cited three times or more are software vendors or marketing agencies. The research firms founders quote in board decks were not cited once.

Why AAJ runs this index

AAJ's resource hub rests on a simple bet: if the answer to a founder's marketing question is specific, sourced and easy to quote, AI answer engines will cite it. Bets need a baseline. This index asks the engines the same 40 questions on each run and records what they cite, so AAJ can see whether that bet is paying off for AAJ and for anyone else writing for founders.

This is the first edition. AAJ does not appear in it at all, and that result is published too.

Search behaviour is the first filter

Perplexity searched and cited on all 40 prompts. Claude cited sources on 10 of 40: four of those were AI search questions, where the answer changes month to month, and the rest were paid media (2), retention (2), go-to-market (1) and sales pipeline (1). GPT-4.1 cited sources on 2 of 39.

AAJ's read: for evergreen "how do I" questions, two of the three engines lean on what they learned in training. Being cited today is only half of visibility. The other half is being part of what the models already know, which rewards publishing consistently over time rather than chasing one query.

There is no dominant source

The 241 citations went to 178 domains. The most-cited domain earned 6, which is 2.5% of the total, and the top ten together took 17%. On a concentration scale where 1.0 means one source takes everything, this set scores 0.008.

For a founder or a small firm, that means the field is open. A clear, specific page can be cited alongside large vendors, because the engines are not defaulting to a handful of publishers.

Vendors and agencies get cited, research firms do not

Of the 14 domains cited three times or more, 10 are software vendors or marketing agencies writing about their own category. The others are LinkedIn, a venture capital firm, a trade publication and a training company.

Gartner, Forrester, McKinsey, Harvard Business Review, Bain, BCG and the other research sources on AAJ's list received none of the 241 citations. University and academic sites received 4. Much of their research sits behind registration walls, which may be why, but this study cannot tell.

AAJ is not in the answers yet

AAJ was named in 0 of 119 answers and cited in 0. With this many answers, that means the true rate is below 3% at 95% confidence. The hub is new, and the next edition, in the first quarter of 2027, will show whether it moves.

By engine

The most-cited domains (three citations or more)

By topic (all three engines)

How the index was run

Prompts: 40 questions, 5 in each of 8 topics that match the AAJ resource hub, worded the way a founder would ask. None mentions AAJ. The set stays fixed between editions so the results are comparable.

Engines: Perplexity through its Agent API (fast preset, the successor to Sonar), Claude through the Messages API (claude-sonnet-4-5 with the web search tool) and OpenAI through the Responses API (gpt-4.1 with the web search tool). Each engine decided for itself whether to search.

Counting: every cited URL is reduced to its registrable domain. AAJ presence is a whole-word match on the answer text; AAJ citation is a link to aajconsult.com. Every rate is shown with its count and a 95% Wilson interval.

Capture: 28 September 2026, run from AAJ's own machine with AAJ's own API keys. The capture code and the prompt set are in the AAJ repository, and the per-answer citation data is a free download on this page.

What this index cannot tell you

• These are API answers, not the consumer apps. ChatGPT, Claude.ai and Perplexity's own apps can search more or less often than their APIs, and the models behind them change.
• One capture on 28 September 2026, one answer per prompt per engine. The same prompt asked again can produce a different answer and different sources.
• Forty prompts per engine is a small sample. The intervals on this page are wide on purpose: a gap between two engines is only a finding when the intervals barely overlap.
• One OpenAI answer failed with a rate-limit error and is excluded, so OpenAI's denominator is 39.
• A domain counts once per answer however often it is linked. For Perplexity, the index counts the sources the answer cites inline, or every listed source when an answer carries no inline markers.
• The labels in the most-cited table (vendor, agency and so on) are AAJ's classification of each company. The research list is fixed in the kit and published with it.

To run the same measurement for your own brand, use the free AI Visibility Significance Test or give the geo-citation-tracker skill to your coding agent. For whether an agent can act on your site once it finds you, see the Agent Readiness Study 2026.

The 40 prompts

1. How do I write a positioning statement for a B2B SaaS startup? (Strategy & Positioning)
2. What is the difference between positioning and messaging for a startup? (Strategy & Positioning)
3. How do I choose an ideal customer profile when we sell to several industries? (Strategy & Positioning)
4. How can a small startup differentiate in a crowded software category? (Strategy & Positioning)
5. What questions should I ask customers to test our positioning? (Strategy & Positioning)
6. Which marketing channel should a seed-stage B2B startup start with? (Go-to-Market & Growth Planning)
7. How do I build a 90-day marketing plan for a startup? (Go-to-Market & Growth Planning)
8. When should a startup hire its first marketer? (Go-to-Market & Growth Planning)
9. Is a fractional CMO worth it for a Series A startup? (Go-to-Market & Growth Planning)
10. How do I plan a B2B product launch with a small team? (Go-to-Market & Growth Planning)
11. How should an early-stage SaaS company set its price? (Pricing & Monetization)
12. Seat-based or usage-based pricing: which is better for a B2B SaaS startup? (Pricing & Monetization)
13. How do I run a price increase without losing customers? (Pricing & Monetization)
14. What should a SaaS pricing page include? (Pricing & Monetization)
15. How do I use the Van Westendorp price sensitivity method? (Pricing & Monetization)
16. How should a startup split its paid advertising budget across channels? (Paid Media)
17. Is LinkedIn advertising worth it for early-stage B2B startups? (Paid Media)
18. How do I know if my Google Ads are profitable? (Paid Media)
19. How do I measure whether paid ads actually caused sales? (Paid Media)
20. What is a good cost per lead for B2B SaaS? (Paid Media)
21. How do I get my startup mentioned in ChatGPT answers? (AI Search & Agent Readiness)
22. What is generative engine optimization and how is it different from SEO? (AI Search & Agent Readiness)
23. Does llms.txt help a website appear in AI answers? (AI Search & Agent Readiness)
24. How do I track whether AI assistants recommend my brand? (AI Search & Agent Readiness)
25. What content gets cited by AI search engines? (AI Search & Agent Readiness)
26. What percentage of revenue should a startup spend on marketing? (Analytics, Experiments & Budget)
27. How do I calculate CAC payback period? (Analytics, Experiments & Budget)
28. What is a good LTV to CAC ratio for SaaS? (Analytics, Experiments & Budget)
29. How do I build a marketing dashboard for a startup board? (Analytics, Experiments & Budget)
30. How do I run a geo holdout test for marketing? (Analytics, Experiments & Budget)
31. How do I calculate net revenue retention? (Retention, Expansion & Advocacy)
32. How do I build a cohort retention analysis for a SaaS product? (Retention, Expansion & Advocacy)
33. What causes early churn in B2B SaaS and how do I reduce it? (Retention, Expansion & Advocacy)
34. How do I design customer onboarding for a B2B product? (Retention, Expansion & Advocacy)
35. What is a good logo churn rate for an early-stage SaaS company? (Retention, Expansion & Advocacy)
36. How do I forecast sales pipeline for a startup? (Sales & Pipeline)
37. How much pipeline coverage do I need to hit a quarterly target? (Sales & Pipeline)
38. How should a startup run win-loss interviews? (Sales & Pipeline)
39. How do I build a target account list for ABM? (Sales & Pipeline)
40. How do marketing and sales hand off leads in a small B2B company? (Sales & Pipeline)

Data

One row per engine and prompt, with every cited URL. Free to reuse with credit to AAJ, AI Answer Visibility Index 2026-Q4 (CC BY 4.0). Download the CSV.