The job description has changed. A communications leader at a Fortune 500 firm now has to answer a question that didn't exist three years ago: when a buyer asks ChatGPT about our category, do we show up — and what does it say about us?
This is not a marginal concern. Reuters Institute survey work tracks the rapid migration of search behavior toward AI answer engines, with publishers forecasting steep declines in traditional search referrals over the next three years. ChatGPT alone reports hundreds of millions of weekly active users, a meaningful share of whom now use it for product research, vendor comparison, and category discovery. The same shift is visible in Google's AI Overviews rollout, which now intercepts a large share of informational queries before users ever click a result.
The discipline forming around this shift goes by various names — Generative Engine Optimization (GEO), AI search optimization, LLM visibility — but they describe the same function. Take the inputs LLMs use to construct answers (training data, retrieval-augmented sources, indexed content, brand entities, third-party citations), and shape them so that the brand surfaces accurately and favorably when a relevant query is asked. The output is no longer a press clipping. It is a citation in a generative answer.
communications, not SEO
There is a temptation to file GEO under search engine optimization and hand it to the digital agency. That is a mistake. The inputs that shape AI answers are heavily weighted toward earned media, owned thought leadership, structured brand entities, Wikipedia, and authoritative third-party content. Those are communications outputs. The skill set required — narrative architecture, source cultivation, message discipline, executive positioning — sits in PR, not paid digital.
The agencies and in-house teams currently winning at this are the ones treating GEO as a strategic comms layer rather than a tactical SEO add-on. 5W, among others, has built a dedicated practice that combines traditional earned media work with proprietary research into how brands surface across major LLMs. The methodology matters less than the posture: the function reports into communications.
What GEO work actually involves
Stripped of vendor pitch language, the work breaks into four streams.
Audit. Run a defined set of brand and category queries across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. Capture how the brand appears, what sources are cited, what sentiment is conveyed, and what gaps or inaccuracies exist. This is the baseline.
Source cultivation. Identify which third-party sources the LLMs are pulling from and prioritize earned media placements that reinforce or correct the picture. A favorable Forbes profile that addresses a common buyer concern is worth more than a top-of-funnel mention in a low-authority outlet — because the LLMs read authority in roughly the same hierarchy that journalists do.
Owned content architecture. Brand newsrooms, executive thought leadership, FAQ structures, and structured data all become more important. Schema.org markup and clear topical authority pages help retrieval-augmented systems find the brand's own answers when relevant.
Entity hygiene. Wikipedia, Crunchbase, Wikidata, and the broader knowledge graph layer feed brand entity recognition across most major models. These are not SEO assets — they are PR territory that requires careful, policy-compliant attention.
The measurement problem
The honest answer is that AI visibility measurement is in its early innings. There is no equivalent yet of Cision-style media monitoring with universal acceptance. Most of the available tooling is either a wrapper around manual querying or a synthetic benchmark of dubious external validity. The teams doing this seriously are building their own query libraries, running them on a regular cadence, and reporting trend data rather than single-snapshot scores.
This will change. AMEC's Barcelona Principles 4.0 update reflects an industry-wide push toward outcome-based measurement, and AI citation behavior is a logical extension. Expect a measurement standard to emerge within 18 to 24 months. Until then, the work proceeds on a defensible-but-imperfect basis — which is how earned media measurement worked for most of its history anyway.
What to do
Communications leaders who want to move now should commission a baseline audit, identify the three to five queries that matter most to revenue, and assign explicit accountability for AI visibility to a named person on the team. The discipline will mature. The brands that began the work earliest will compound an advantage that compounds the same way SEO authority compounded in 2008 to 2012 — slowly, then suddenly.
The function does not need a big budget. It needs a clear owner.
