Were we named?
Did the assistant attribute the news to us, or describe it anonymously?
Insight · Strategy
How AI Search Changed PR in 2026
Fewer clicks, more summarised answers, and a new question for comms teams: if the assistant answers without linking, did your announcement land?
For twenty years, the PR measurement question was whether your link got clicked. Increasingly, the answer is that there was no link and no click — the assistant simply told the user what you announced, and may or may not have named you.
Quick answer
AI search reduced referral clicks while increasing the number of places an announcement can be repeated. Comms teams now need to optimise for attribution and accuracy inside summarised answers, not only for traffic, and to measure citation share alongside referrals.
Key takeaways
- Referral traffic from search is declining for informational queries.
- Being named in an answer is now a distinct outcome from being clicked.
- Accuracy of what a model says about you is a comms responsibility, not an SEO one.
- Distribution breadth is the main lever for corroboration, which drives citation.
What shifted
Three changes arrived close enough together to be mistaken for one:
- Answers replaced result lists. For factual and comparative questions, a generated summary now answers without a click.
- Retrieval replaced recall. Assistants increasingly fetch live sources rather than relying on training data, which makes current, well-structured pages matter again.
- Attribution became inconsistent. Some engines cite sources prominently, others do not. The same question can produce a cited answer on one platform and an uncited one on another.
The click decline
If your PR reporting still leads with sessions, it will show a decline that is not a failure of the campaign. The audience did not disappear; the interaction changed from a click to a summary.
That does not make referrals irrelevant — referral quality from earned coverage remains one of the strongest signals available. It means referrals are no longer a complete scoreboard.
New questions for comms teams
Were we accurate?
Did it state the right figures, product, and organisation?
Which source did it use?
Our release, a syndicated copy, or a competitor's coverage?
Who was named instead?
If a competitor appeared in our category answer, why?
A practical playbook
- Publish releases as clean, indexable pages with structured data — not PDFs or images of text
- Put a standalone answer block in the first 60 words of every release
- Say your organisation's name the way you want it repeated, every time
- Get the same facts into independent outlets so they can be corroborated
- Track a fixed question set across engines on a schedule, and record accuracy as well as presence
- Publish corrections quickly — a model repeating a stale error is a long-term problem
The uncomfortable implication: PR output that exists purely to generate links is worth less than it was. PR output that establishes facts clearly, and gets them repeated, is worth more.
Our AI Visibility Suite applies this to every release we distribute, and the AEO, GEO and LLMO guide covers the mechanics.
undefined
undefined
undefined
undefined
Get started
Put the guide into practice.
Submit a release and our editors will apply these standards to it — including telling you what they would cut.
No subscription required · Cancel anytime · Human editorial review on every release