Search & AI Visibility: The Complete Guide to Being Found in Google, ChatGPT and AI Search

Being "found" used to mean one thing: a page appearing somewhere on a Google results page. It now means several different things at once — a page ranking in Google, a passage being summarised in an AI Overview, a business being mentioned by ChatGPT, Gemini, Perplexity or Copilot, and, separately, whether any of that actually produces an enquiry. These are related systems, not the same system, and treating them as one is the most common reason B2B companies misjudge their own visibility.

This guide sets out, in plain English, how traditional search and AI search actually work, what a business can reasonably influence and what it cannot, and a practical, phased route to improving visibility across both — without pretending that any of it can be guaranteed. It is the pillar for a series of more specific articles; where a topic has its own detailed answer, this guide names it and points to it rather than repeating it in full.

Two systems, one set of foundations

Traditional search and AI search are frequently discussed as though they are competitors, or as though one is replacing the other. Neither is quite right. Google is adding AI-generated answers (AI Overviews, AI Mode) on top of its existing index, and OBSERVATION suggests buyers increasingly also ask standalone AI assistants directly. But both routes depend on broadly the same underlying material: a website that is technically accessible, that states clearly what the company does, for whom and where, and that is corroborated by independent sources. A business that gets those foundations right improves its chances in both systems at once; a business that neglects them is working against itself in both.

How traditional search works: crawl, index, rank

Google's process is documented in three stages. Crawling: Google's automated software finds pages, generally by following links, and requests them. Indexing: Google analyses a page's content and signals (text, images, structured data) and stores it, if eligible, in its index — the dataset Search is actually drawn from. Ranking: when someone searches, Google selects and orders pages from the index that it judges relevant and useful to that specific query, using hundreds of signals. A page that is not crawled cannot be indexed; a page that is not indexed cannot rank, regardless of how good it is.

FACT: Google documents this crawl–index–rank model directly (Search Essentials, Search Central documentation). It also documents that it renders JavaScript, but in stages — an initial crawl of the raw HTML, then a later rendering pass — so content that depends entirely on client-side JavaScript to appear can be delayed or missed; server-rendered HTML is the more reliable option.

What actually influences ranking

Google does not publish its ranking algorithm, and nobody outside Google knows it in full — that is UNKNOWN by design, not an oversight in research. What Google does document is a set of principles: content should be genuinely helpful and created for people, not primarily to manipulate rankings; pages should be technically sound (crawlable, fast, mobile-usable, secure); and sites should avoid deceptive or manipulative practices. These are documented in Google's Creating Helpful Content guidance and SEO Starter Guide, and they remain the most reliable foundation, because they are the only part of the system Google itself confirms.

How AI search works: training data and live retrieval are different things

This distinction matters more than almost anything else in this area, and it is widely blurred in commentary. A large language model is trained on a fixed snapshot of data up to some cut-off date; what it "knows" from training is frozen until the model is retrained. Separately, many AI products now also retrieve live information at the moment of a query — browsing the web, checking a search index, or calling a specific tool — and use that fresh material alongside or instead of what was learned in training. A business can appear in an AI answer because it was part of the training data, because it was retrieved live, or both, and these have different practical implications.

FACT, by provider, documented in official developer documentation: OpenAI operates separate, controllable crawlers — GPTBot (training), OAI-SearchBot (surfacing in ChatGPT search) and ChatGPT-User (fetches triggered by a live user request) — and states that sites opting out of OAI-SearchBot will not be shown in ChatGPT search answers, though they may still appear as plain navigational links (platform.openai.com/docs/bots). Perplexity operates PerplexityBot for indexing and Perplexity-User for user-triggered fetches (docs.perplexity.ai/guides/bots). Microsoft documents that Copilot grounds its answers using Bing's search results, which makes being indexed by Bing relevant to Copilot visibility, though the detail of how Copilot selects and weights sources beyond that is not public.

Citations, mentions and recommendations are not the same event

• A citation or source link: the AI answer explicitly references a specific page as the source of a claim.
• A mention: the business's name appears in the answer text, without necessarily linking to a specific page.
• A recommendation: the AI presents the business as a suggested option in response to a buying-type question (e.g. "who supplies X").

OBSERVATION: practitioners report that these three outcomes behave differently and are influenced by different combinations of factors — clarity of the underlying content, third-party corroboration, and how the specific question is phrased. This is not formally documented by any provider, so treat it as a reasonable operating distinction rather than a proven mechanism.

Authority: why being named elsewhere matters

Authority, in both traditional and AI search, is less about any single ranking factor and more about whether independent sources corroborate what a company says about itself. A website that claims expertise is weaker evidence than a website whose claims are echoed — consistently — by customer reviews, press coverage, directory listings, industry bodies and other sites. This is consistent with how search engines are documented to treat trust signals generally, and is a reasonable INFERENCE for how AI systems weigh competing or unverified claims, though no provider publishes the exact mechanism. The practical implication is the same regardless: consistency of the basic facts (what you do, for whom, where) across your own site and everywhere else you're mentioned is worth more than any single piece of content.

Content: answering real buyer questions, specifically

Both systems reward content that answers a specific question clearly, in a self-contained way, rather than content written primarily to include keywords or to pad out a page. For a B2B company this usually means: a dedicated, specific page per service and per significant question a buyer actually asks, written in plain language, with the direct answer near the top rather than buried under preamble. It does not mean publishing a high volume of thin articles — see the related question of how much content a B2B site actually needs, and whether 100 articles helps or dilutes a site's authority, covered in its own article in this series.

Website architecture: service and location pages

A website that puts a dozen services on a single page, or splits one service across three similarly worded pages, gives both Google and AI systems an ambiguous signal about which page is authoritative for that topic — a problem known as keyword cannibalisation, covered separately. The practical fix is architectural: one clear, specific page per service, one clear page per location served (where location genuinely matters commercially), internal links connecting related pages, and a consistent statement of what the company does repeated across the site rather than reworded each time. This overlaps directly with how a B2B site should be structured for growth generally — a question this guide does not re-answer in full, because a dedicated article already does.

Entities, in plain English

Entity

In search terminology, an entity is simply a specific, identifiable thing — your company, a particular service, a location, a person — that a search or AI system can recognise as distinct from similar things, rather than a vague reference it has to guess at. Making your company an unambiguous entity means it is always clear, across your own site and anywhere else you are mentioned, who you are, what you sell, where you operate and which page is the authoritative one for each service.

This matters practically because inconsistency — a company named three slightly different ways, a service described differently on different pages, conflicting location information — forces a search or AI system to guess, and systems that have to guess are less likely to cite or rank you with confidence.

Structured information (schema)

Structured data (schema.org markup) is a way of stating facts about a page in a format machines can read directly, rather than requiring them to infer the facts from prose. FACT: Google documents that structured data must accurately reflect visible page content, and that it can make a page eligible for certain rich results — it does not guarantee rankings, and since 2023 Google has restricted FAQ rich results in search results to well-known authoritative government and health sites. Whether and how much structured data influences AI citation is UNKNOWN and not published by any AI provider; it is reasonable practice, not a guaranteed lever. A dedicated article in this series covers this in full.

Measurement: five things that are not the same

Limitations and uncertainty: what nobody can promise

• No one outside Google, OpenAI, Microsoft or Perplexity knows the full ranking or retrieval mechanics, and all of them change without announcement.
• No content format, schema type, or volume of articles guarantees a citation, ranking position or AI recommendation.
• Allowing an AI crawler does not guarantee inclusion in that platform's answers; blocking one does not guarantee exclusion from everything the platform does.
• AEO and GEO (answer-engine and generative-engine optimisation) are largely industry labels describing good SEO and content practice applied to AI surfaces, not a separate, proven technical discipline — covered in full in the next article in this series.
• No agency, including Evans, can guarantee a Google ranking position or an AI mention for any specific query.

A practical implementation roadmap

Most B2B companies improve visibility fastest by working in phases, rather than trying to fix everything simultaneously. The following is a reasonable sequence, not a fixed formula:

1. 01Foundations: a crawlable, server-renderable site; correct robots.txt and sitemap; no accidental blocking of search or AI crawlers; fast, mobile-usable pages. Diagnosed properly in a Commercial Website Audit.
2. 02Clarity: one specific page per service and per location that genuinely matters commercially, each stating plainly what it covers, for whom, and why the business is credible; a consistent description of the company repeated across the site rather than reworded on every page.
3. 03Authority: content that answers real buyer questions specifically, built up deliberately rather than in volume for its own sake, plus genuine third-party corroboration — reviews, press, directories, industry mentions — kept consistent with the website's own claims.
4. 04Measurement: tracking indexing, ranking, traffic, AI mentions (manually, since no platform reports this automatically) and — the part that actually matters commercially — enquiries, so that visibility work is judged by leads and customers, not by vanity metrics.

Where this fits commercially

For a B2B business, none of this is worth doing for its own sake. The point of search and AI visibility is being found by the right buyer at the point they are deciding who to talk to — and then converting that visibility into an enquiry. A company that ranks well but converts poorly, or that is occasionally mentioned by ChatGPT but has no clear next step for the reader, has not solved its commercial problem. Visibility, content, website structure and conversion are one connected system, which is why Evans treats them together rather than as separate technical projects.