AI visibility #
How easily AI assistants such as ChatGPT, Gemini, Perplexity, and Claude can find, understand, and recommend a business when someone asks a relevant question. Unlike search rankings, AI visibility is invisible to the business itself: there is no public results page, only the answers other people receive. It is built from readable pages, consistent business data, structured markup, and mentions in sources assistants trust.
In padel: A padel club with strong AI visibility gets named when a player asks ChatGPT where to play tonight.
GEO (generative engine optimisation) #
The practice of improving how often and how favourably generative AI systems mention a business in their answers. GEO overlaps with SEO because both reward clear content and structure, but the output differs: SEO produces a position on a ranked list of links, GEO produces a mention inside a single generated answer, where the AI chooses on behalf of the user.
In padel: GEO work for a padel brand aims to get it named when assistants answer which racket to buy.
AEO (answer engine optimisation) #
A near-synonym of GEO, framed around answer engines: systems like Perplexity or Google AI Overviews that reply to questions with one composed answer instead of a list of links. AEO focuses on becoming the source that answer engines quote and cite, typically by publishing direct, well-structured answers to the exact questions people ask.
In padel: AEO for a padel academy means owning the answer to padel lessons for beginners in its city.
llms.txt #
A plain text file placed at the root of a website that describes the site for AI systems: what the business is, who it serves, and a curated list of its most important pages with one-line descriptions. It is an emerging convention rather than an enforced standard, cheap to implement, and useful as a machine-readable orientation to your site.
In padel: A padel club's llms.txt would point AI systems to its booking, prices, courts, and coaching pages.
Share of voice (AI) #
The percentage of brand mentions a business captures when AI assistants answer the buying questions in its category. It is measured by asking a fixed panel of realistic questions repeatedly across platforms and counting every brand named in the answers. Share of voice turns AI visibility into a comparable, trackable number with a competitive baseline.
In padel: Our racket report measures each padel brand's share of voice across OpenAI, Perplexity, and Gemini in Spain.
AI crawlers #
The automated agents AI companies use to read the web, either to gather training data or to fetch pages live while answering a question. Each announces itself with a user agent name, and websites allow or block them in robots.txt. Blocking AI crawlers makes a business invisible to the assistants that rely on them.
In padel: A padel club that blocks AI crawlers in robots.txt cannot be read, and so cannot be recommended.
JSON-LD #
The standard format for adding structured data to a web page: a small script of labeled facts, invisible to visitors, that states in machine-readable form what the page and the business are. Search engines recommend it and AI systems parse it. JSON-LD does not change a page's design; it removes the guesswork from how machines interpret it.
In padel: A padel venue's JSON-LD can declare its address, opening hours, court count, and booking URL.
FAQPage schema #
A structured data type that mirrors the visible questions and answers on a page in machine-readable form. Because people phrase prompts to AI assistants as questions, FAQ content maps directly onto real queries, and FAQPage markup helps machines connect each question to its answer. The markup must match the visible text; schema for answers that do not appear on the page erodes trust.
In padel: A club FAQ on whether beginners can rent a racket, marked up correctly, matches a common padel prompt word for word.
Grounded search #
An AI answering method where the system retrieves live web pages first, then writes its answer from what those pages say, usually with citations. Grounding reduces invented facts and makes the answer as current as the retrieved sources. Perplexity and ChatGPT's browsing mode work this way, which is why fresh, readable pages can influence answers within weeks.
In padel: A grounded answer about padel courts in Lisbon quotes and cites the venue pages it just read.
Memory-mode AI #
An AI assistant answering from what its model already learned during training, without consulting the live web. Memory-mode answers reflect the web as it looked when the training snapshot was taken, so recent changes are absent and long-standing, widely corroborated facts dominate. Businesses influence memory mode slowly, by being clearly and consistently documented before the next training round.
In padel: In memory mode, an assistant may recall a padel club that closed last year and miss one that opened last month.
Citation #
A source reference an AI answer points to, identifying the specific page a claim was drawn from. Citations are the visible currency of answer engines: being cited means your page directly shaped the answer and earns a clickable link inside it. Watching which pages get cited for your market's questions reveals your real competitive set.
In padel: When Perplexity answers a padel racket question, each numbered citation is a page that won a seat in the answer.
Entity #
A distinct thing machines can recognise and track across the web: a business, a person, a place, a brand. Machines resolve entities by matching names, addresses, and identifiers across many sources; consistency strengthens the entity, contradictions weaken it. A business whose name differs across its website, map listing, and directories can look like several weak entities instead of one strong one.
In padel: A padel club using one identical name everywhere becomes a single entity assistants can confidently recommend.
E-E-A-T #
Experience, Expertise, Authoritativeness, Trustworthiness: the quality framework search engines use to judge content and its sources, increasingly relevant to what AI systems cite. Content demonstrating first-hand experience, named credentialed authors, and corroborated claims tends to be preferred over anonymous or derivative text. For businesses, E-E-A-T mostly means showing verifiable evidence of being what you claim to be.
In padel: A padel coaching guide written by a named, federation-certified coach carries E-E-A-T that anonymous content lacks.
Knowledge graph #
A structured database of entities and the relationships between them, used by search engines and AI systems as a backbone of verified facts. When a business exists in knowledge graphs with correct details, machines answer questions about it confidently. Entries are built from authoritative sources: official sites, structured data, map listings, directories, and encyclopedic references.
In padel: A padel chain in the knowledge graph gets its venues, cities, and opening years stated as facts, not guesses.
Hallucination #
A confident but false statement produced by an AI system, typically when it lacks reliable information and fills the gap plausibly. For businesses, hallucinations show up as wrong addresses, invented prices, or misattributed offers. The practical defence is publishing clear, corroborated facts: assistants hallucinate least about entities that are thoroughly and consistently documented.
In padel: An assistant inventing an eight-court indoor hall for a padel club that has four outdoor courts is hallucinating.
Prompt #
The text a person types to an AI assistant: the question or instruction the answer responds to. Prompts are the query data of the AI era, and their phrasing tends to be longer and more conversational than search keywords. Understanding the real prompts customers use reveals which questions a business needs to answer publicly to be part of the response.
In padel: A padel court for four beginners tonight with rental rackets is a prompt, and a bookable answer to it wins.
Answer engine #
A system that responds to a question with one composed answer instead of a list of links: Perplexity, ChatGPT with browsing, Google AI Overviews. Answer engines compress many sources into a few sentences, so far fewer businesses get surfaced per query than on a traditional results page. Winning placement in the answer replaces ranking as the objective.
In padel: An answer engine asked about padel in Milan names two or three venues; there is no page two.
Structured data #
Information published in a standardised, machine-readable format, most commonly schema.org vocabulary embedded as JSON-LD. Structured data lets a website state facts directly (business type, location, hours, offers, events) instead of leaving machines to infer them from prose. It is one of the highest-leverage technical fixes in AI visibility because it removes ambiguity at the source.
In padel: Structured data on a padel tournament page declares its date, venue, and registration link machine-readably.
Canonical URL #
The single authoritative address a page declares for itself, telling crawlers which version to index when the same content is reachable at several URLs. Canonicals concentrate the signals for a page in one place and prevent machines from splitting attention across duplicates. Every indexable page should declare one, pointing at the version you want machines to treat as real.
In padel: A padel club's prices page reachable with and without tracking parameters should declare one canonical URL.
Sitemap #
A machine-readable file, usually XML, listing the pages of a website so crawlers can discover them all reliably. Sitemaps do not improve rankings by themselves; they guarantee findability, including for new or deeply linked pages, and communicate when pages change. Referenced from robots.txt, they are part of the basic plumbing of being fully crawlable.
In padel: A padel chain's sitemap lists every venue page, so a new club's page is discovered the week it launches.