There’s an awful lot going on in the world at the moment. Everywhere is either immersed in conflict or appears to be on the brink of war, so mundane matters like practice notes and blog posts seem somewhat trivial.
But we still need a bit of business as usual and I hope this one gives you some food for thought as you wrangle your organisation and clients through the current environment, helping them demonstrate the value of what they do and what you do.
I first tackled public relations measurement and evaluation way back in the last century when, at the now CIPR, we were pulling together first guidelines on the topic under the guidance of David Phillips. Since then we’ve made slow but steady progress, moving away from simply counting the stuff we send out and towards understanding the nature of organisational relationships, supporting strategic outcomes, deepening social capital and maintaining our licence to operate.
And then AI arrived and, apparently, we decided doing the time warp and going back to counting stuff out would be fun.
If you haven’t yet encountered AEO, AIO or GEO, you soon will. Answer Engine Optimisation, AI Overview Optimisation and Generative Engine Optimisation are the latest attempts to make organisations more visible when people ask questions through AI-powered search and generative systems.
That sounds reasonable enough.
If somebody asks Google, ChatGPT, Claude or another system a question relevant to your organisation, of course you want the information they receive to be accurate, current and useful.
If somebody asks Google, ChatGPT, Claude or another system a question relevant to your organisation, of course you want the information they receive to be accurate, current and useful.
The trouble starts when visibility itself becomes the prize.
A new industry has sprung up around measuring how often organisations appear in AI-generated answers, which models mention or cite them, how prominently they feature and how they compare with competitors. Alongside the alphabet soup come dodgy dashboards, proprietary indices, unexplained visibility scores and persuasive promises of competitive advantage.
A new industry has sprung up around measuring how often organisations appear in AI-generated answers, which models mention or cite them, how prominently they feature and how they compare with competitors. Alongside the alphabet soup come dodgy dashboards, proprietary indices, unexplained visibility scores and persuasive promises of competitive advantage.
Some early adopters are understandably cock-a-hoop about all this. Being cited by an AI system feels like being discovered. A rising visibility score looks like progress.
But what, exactly, has happened?
An AI-generated answer is not a fixed search result. Ask different systems the same question and you can receive completely different answers. Ask the same system again and the sources, framing and conclusions may change. Citations do not necessarily support the claims attached to them and the systems can hallucinate, draw from poor-quality material or confidently blend fact and fiction.
Then vendors place their own scoring systems over the top.
If somebody tells you your 'AI visibility score' is 68, resist the applause and ask the obvious question.
68 what?
What has been counted? What has been weighted? Which prompts were used? Which models? Over what period? And why should that proprietary concoction be considered evidence of success?
Useful observations can certainly be made. We can test whether an organisation appears in seven out of ten searches, identify which sources are being used, find inaccuracies and see how the organisation is framed.
That is intelligence, not an outcome. And there is a bigger issue still.
Traditional search at least offered a recognisable journey. Somebody had a question, searched for information, found a source and went to the organisation, publisher or information provider to learn more.
Generative search changes that journey because the machine now sits in the middle. It retrieves information, selects it, interprets it, combines it and presents its own answer before the person necessarily encounters the original source at all.
In Google’s AI Overview environment and similar systems, the answer may be sufficiently complete that the person never leaves the platform.
So here is the paradox.
Your organisation may become more visible to the machine while becoming less connected to the person. That is why I think we need to look beyond the new measurement illusion and consider the illusion of communication itself.
We might be cited. We might be mentioned. We might even be recommended.
But a mention does not demonstrate knowledge. A citation does not demonstrate trust. A recommendation does not demonstrate commitment. And none of them tells us whether the person understands us, believes us, wants a relationship with us or will ever visit our own information environment.
For public relations practitioners, that is the real issue.
Our work is concerned with building and sustaining the relationships organisations need to keep their licence to operate. Those relationships depend on trust, satisfaction, mutuality, loyalty, commitment, reputation and understanding.
So we absolutely should monitor this new environment. Test the high-intent questions stakeholders and communities are likely to ask. Compare answers across systems. Check sources. Look for missing information, bad framing, hallucinations and emerging reputational risks. Audit what your organisation makes available to both people and machines.
But don’t confuse being visible to an algorithm with being connected to a human being.
That distinction is becoming increasingly important.
It’s also the territory I’ll be exploring in Now You See Me, Now You Don’t – AEO, AIO, GEO and the New Measurement Illusion, our live session on 13 October.
Because before we start celebrating AI visibility, we need to understand what has become visible, to whom, through whose system and what happened to the relationship along the way.