Assistants are describing your business to buyers, daily.

Because a growing share of buyers ask them first, and the answers are consequential in ways you cannot see in analytics: which competitors get shortlisted alongside you, whether your pricing or service scope is described accurately, whether an outdated fact from a five-year-old page is being repeated as current, and whether you appear at all for the questions that precede purchase. Businesses routinely discover, through monitoring, that assistants are recommending competitors for queries they dominate in classic search.

By prompt sampling, because no engine publishes rank data. We define the prompt set that matters commercially (your money questions, category shortlists, comparison queries, brand-specific questions), run them on a schedule across ChatGPT, Perplexity, Claude, Copilot and Google AI Overviews, and log presence, sentiment, accuracy and which sources get cited. Results vary between runs by design, so trends across samples matter more than any single answer, and we report it that way.

Trace and correct at source: assistants repeat what the corroborated web says, so misinformation almost always originates somewhere findable, a stale directory listing, an old press release, an outdated page you forgot, a competitor comparison, or an unclaimed profile. The fix is correcting those sources and strengthening the accurate ones, then re-sampling to confirm drift. There is no support desk for LLM corrections; the substrate is the web, which is why entity hygiene is the actual remedy.

Influence yes, control no, and the distinction matters commercially. Influence comes from being unambiguous and corroborated: consistent entity data, structured markup, clear factual pages, presence on the third-party surfaces engines lean on, and content that answers the questions being asked in extractable form. Anyone claiming deterministic control over model outputs is describing something that does not exist.

AI SEO is the building work (citability engineering, structure, entity consistency, corroboration); monitoring is the measurement and early-warning layer that tells you whether it worked and what changed. Most clients buy them together, monitoring standalone suits businesses with in-house teams who want the intelligence without the implementation, or brands whose main concern is reputation accuracy rather than acquisition.

From £450/month for a defined prompt set sampled across engines with monthly reporting and alerting on material changes. Larger prompt sets, multi-market or multi-language sampling, and competitor benchmarking scoped higher. One-off baseline studies from £600 when you want to know where you stand before deciding whether ongoing tracking is worth it.