Branch surveyed 300 enterprise marketing, growth and digital leaders for its 2026 AI Search and Discovery benchmark. The reported results show organisations moving quickly, but they also expose a gap between confidence and measurement.
The most useful conclusion is not that every business should redirect budget into a new channel. It is that customer discovery is becoming harder to separate into neat “SEO,” “AI” and “direct” boxes, while many organisations still lack the evidence to attribute the whole journey.
The reported findings
Branch's report landing page publishes three headline results:
- 87% of respondents expected AI platforms to complete transactions for their company within 12 months.
- 66% said they were confident in AI attribution, yet 26% could not track the customer journey from AI discovery to conversion.
- 28% planned to dedicate more than half of their 2026 marketing budget to AI search optimisation.
A sponsored summary written by Branch's Head of Strategic Growth adds another budget result: 65% reported allocating at least a quarter of marketing budget to AI. The same summary says fewer than one in five reported no AI measurement challenges.
These numbers are statements from the surveyed executives. Public materials reviewed for this article do not provide enough methodological detail to treat them as representative of all enterprises, South African companies, or realised financial performance.
Finding 1: expectations are moving faster than proof
Expecting transactional AI agents within a year is strategically interesting. It is not evidence that a company's customers already use those journeys at scale.
Before building or buying for agentic transactions, verify:
- whether customers use AI tools during discovery or comparison;
- which website actions an agent could safely complete;
- whether prices, availability, policies and service data are structured and current;
- how consent and personal data would be handled; and
- what human fallback is required when the request is ambiguous.
This separates readiness work from speculation.
Finding 2: confidence can conceal attribution gaps
The apparent tension between 66% confidence and 26% inability to trace the journey is the most actionable result. A referral from an AI platform may capture the last visible step while missing earlier conversations, branded searches, direct visits, retargeting and offline contact.
An honest measurement model therefore has layers:
| Layer | Useful evidence | Limitation |
|---|---|---|
| Direct referral | Source and landing page in analytics | Misses unlinked answers and later visits |
| Search visibility | Search Console pages, queries and AI-feature reports | Does not show every cross-platform journey |
| Conversion | Qualified form, call, booking or sale | Attribution may be shared across channels |
| Customer evidence | “How did you hear about us?” and sales notes | Self-reporting is imperfect |
| Incrementality | Controlled geographic, audience or time-based tests | Requires enough data and careful design |
No single row solves attribution. Together they make overconfident claims less likely.
Finding 3: budget percentages are not a benchmark
The reported allocations are unusually high for an emerging channel. A company should not copy them without knowing the survey sample, definition of “AI search optimisation,” current channel economics and internal capabilities.
For a South African SME, a safer order is:
- fix crawlability, indexing, analytics and high-intent pages;
- identify real AI referrals and customer mentions;
- test a small prompt and citation sample;
- improve content and data that support both search and customer decisions; and
- expand investment only when qualified outcomes justify it.
Much of this work is still good SEO, content quality and measurement rather than a separate budget line.
Finding 4: AI and SEO are usually additive
A person may begin in a chatbot, verify through Google, visit a review site and convert after a branded search. Treating the channels as mutually exclusive creates false attribution and fragmented content.
The common foundation is clear, indexable and reliable information. Google's own generative Search guidance says established SEO practices remain relevant because AI features use the Search index and ranking systems.
Finding 5: reported sentiment is not business performance
The sponsored summary states that only 3% of respondents reported negative marketing performance from AI. This is a perception result from the survey; it does not establish an average return, a causal impact, or the likely outcome for another company.
Use the survey to identify questions for your own measurement, not to fill gaps in your own data.
A practical response
Create one decision-ready dashboard that combines classic Search, generative Search where available, referral traffic, branded demand and qualified conversions. Record unknowns. Avoid assigning revenue twice when more than one platform claims the same conversion.
The strategic advantage is not declaring an AI-search budget first. It is building enough evidence to decide where the next rand should go.
Sources
- Branch: AI Search and Discovery Enterprise Benchmark Report
- Search Engine Journal: Five AI search findings every enterprise marketer needs to know
- Google Search Central: Optimizing your website for generative AI features