Three intents, three sourcing behaviours
AI systems source answers differently for informational, consideration and transactional queries. Researchers at the University of Toronto ran 300 queries, 100 per intent, through four AI systems and Google, and classified every cited source as brand, earned or social, in a study presented at the EDBT/ICDT 2026 workshops. Their summary: source composition shifts systematically by intent.
| Intent | Example | Dominant AI source | Google's mix |
|---|---|---|---|
| Informational | "What is generative engine optimisation?" | Varies by platform: editorial or brand | Balanced, more social |
| Consideration | "Best GEO agencies for EU B2B" | Earned media, 59–86% | Social-heavy, 41% |
| Transactional | "Hire AI visibility agency Netherlands" | Brand pages, 52–68% | Mixed |
The commercial weight sits in the middle row. Consideration queries are where shortlists form, and every AI system tested builds those answers mostly from editorial sources. A brand that is absent from the publications AI trusts for its category is absent from the shortlist, however good its own site is.
What AI cites for informational queries
Informational queries are the least predictable stage: some AI systems lean on editorial explanations, others on brand-owned educational content. The Toronto data shows wide variation between platforms here, with Google alone keeping a balanced mix that includes social threads.
Two things reliably help. Long-form explanatory content performs best across formats: S. K. Iyappan's comparative framework analysis reports a 92% AI citation rate for long-form contextual content, the highest of any format measured. And clarity of structure matters as much as depth, because the assistant extracts passages rather than reading pages. The brand whose explanation of a category is the clearest one available also sets the vocabulary buyers carry into the next stage, which the AI brand influence piece explores.
What AI cites for consideration queries
Consideration queries are answered from earned media in 59–86% of citations across all four AI systems. The Toronto authors describe the platforms as converging on editorial dominance at this stage, in contrast with Google's social-focused approach at 41%.
That has one direct implication: appearing when buyers evaluate vendors is a digital PR outcome, not a website outcome. The target is not high-authority publications in general but the specific outlets Perplexity shows in its citations for your category's consideration questions. Description quality matters too. A mention that states what you do, for whom and with what method gives the assistant usable material; a name-drop does not. The digital PR piece covers how to find and earn those placements.
What AI cites for transactional queries
Transactional queries flip the pattern: all four AI systems sharply increase brand citations, to 52–68%. The buyer who asks "hire an AI visibility agency" is sent to the agency's own service pages, if those pages contain something citable.
That means specifics. Delivery scope, timeline, what a monthly report contains, who it is for. FAQ blocks on service pages with FAQPage schema give the assistant self-contained answers; Iyappan reports 67% citation rates for FAQ-format content against 41% for keyword-focused pages. Omar Haddad's study of AI-mediated commerce found that moving from below-median to upper-quartile structured content completeness raised AI-assisted inclusion by 8.7 percentage points. At the stage where AI already prefers your pages, completeness is the multiplier. See product content optimisation for the page-level checklist.
Building an intent-matched content plan
An intent-matched plan assigns one investment per stage instead of one plan for all three:
- Informational: long-form, well-structured educational content on your own site that frames the category.
- Consideration: editorial coverage in the outlets AI cites for your category, with specific brand descriptions.
- Transactional: complete service pages with operational detail and FAQPage schema.
Measurement follows the same split. Test ten informational, ten consideration and ten transactional queries each month on ChatGPT and Perplexity, and track inclusion rate per stage rather than one blended number. The stage with the widest gap is where the next euro goes. The AI SEO metrics guide covers the routine.
Platform nuance is real but secondary. Claude is the most editorial-dependent system (65% earned overall), Perplexity draws somewhat more on brand pages, and Gemini is the most balanced because it grounds on Google Search. The platform comparison has the detail; the intent split above holds on all of them.
Frequently asked questions
Which intent stage matters most for B2B?
Consideration. It is where vendor shortlists form, and AI systems build those answers from editorial sources at 59–86%. For B2B, editorial coverage in the publications AI trusts for your category is the highest-return single investment.
How do I test my visibility by intent?
Write three lists of ten queries: category education, vendor evaluation, and purchase-intent. Run each in ChatGPT and Perplexity, record whether you appear, at what position and how you are described. Keep the results per stage; a blended inclusion rate hides exactly the gap you need to see.
Do I need separate content for each intent?
Yes. The source types differ, so the content types differ: educational pages for informational, editorial mentions for consideration, structured service pages for transactional. A single "comprehensive" page serves none of the three as well as content built for each.
How is this different from intent in classic SEO?
Classic SEO uses the same intent labels but serves all of them through one mechanism: ranking pages by relevance and authority. AI search changes the source type itself per stage, which is why one-track content strategies underperform there.
Does intent matter as much for a small business?
More, in one respect. At the transactional stage AI cites brand pages regardless of domain authority, so a small firm with complete service pages is on equal footing. At the consideration stage, focusing scarce PR effort on the outlets AI actually cites for your niche beats broad outreach.
Where does social content fit?
Barely. Across all intents the AI systems cited social platforms at 1–11% of sources, against 34% for Google. The consideration stage, where social is strongest in Google, is where AI relies on it least.
Find your widest intent gap
The free analysis shows whether AI assistants name you for your category, and the report's AI section tells you which stage of the buying journey you are missing from.
