Guides.
How AI assistants choose what to recommend, and how to measure it honestly.
01 Fundamentals
- What is AI visibility, and why it now decides who gets chosenAI visibility is how often assistants like ChatGPT, Claude, Gemini and Perplexity name, recommend or cite your brand when buyers ask for options.
- GEO, AEO and SEO: what each means and how they differSEO earns rankings in search results, AEO earns a place in direct answers, and GEO earns mentions in AI-written answers. Here is how they differ and overlap.
- How ChatGPT decides which brands to recommendChatGPT recommends brands using what it learned in training and, when it searches the web, the pages it finds. Here is what shapes the brands it names.
- How Perplexity chooses the sources it citesPerplexity searches the web for every question and cites numbered sources beside its answer. Here is what shapes which pages it reads and which it cites.
- How Gemini answers with Google Search behind itGemini can ground its answers in Google Search results. Here is what that means for which brands it names, and how it differs from classic Google rankings.
- How Claude searches the web before it answersClaude can search the web, read pages and cite sources before it answers. Here is how that works, what it means for brand visibility, and how to test it.
- Why AI answers change every time you askAI answers change between asks because assistants generate text, search live and react to wording and context. Here is why, and how to measure around it.
- Mentioned, recommended or cited: the three ways AI names a brandAI names a brand in three ways: a mention, a recommendation or a citation. Each means something different, and counting them together hides the truth.
- What an AI citation is worth to a businessAn AI citation can bring clicks, trust and a place on the shortlist, but its worth varies. Here is how to judge the value of being cited without inflating it.
- Do AI assistants follow Google rankings?AI assistants partly overlap with Google rankings but do not follow them. Here is where the two connect, where they split, and what that means for your brand.
- Training data or live search: where AI answers come fromAI answers come from two places: knowledge learned in training and, when search is used, pages fetched live. Here is how to tell which one shaped an answer.
- The four AI assistants that matter when people buyChatGPT, Claude, Gemini and Perplexity are the four assistants most worth testing when buyers ask for recommendations. Here is why and how they differ.
02 Checking and measuring
- How to check if ChatGPT recommends your businessTo check if ChatGPT recommends your business, ask the questions buyers ask, without your name, in a clean session, several times, and record every answer.
- How to test AI answers properly: signed out, private, repeatedTest AI answers properly by using a clean signed-out or private session, a fixed question, fresh chats and repeated runs, and by saving every answer in full.
- How many times should you ask an AI before trusting the answer?Ask an AI each question at least several times, in fresh chats, before trusting the result. More repeats give steadier rates, with diminishing returns.
- How to find the questions your customers ask AIFind the questions customers ask AI by mining sales calls, support tickets, search suggestions and forums, then writing them as buyers type them.
- Share of voice in AI answers: what it is and how to measure itShare of voice in AI answers is your brand's slice of all the recommendations across a fixed set of questions. Here is how to measure it and read it fairly.
- How to track AI visibility over timeTo track AI visibility over time, use a fixed question set, ask on a schedule, keep every answer, and compare like with like, judging change with care.
- What a fair AI visibility benchmark looks likeA fair AI visibility benchmark uses a published, versioned question set, no brand names in the questions, repeated runs, kept answers and stated uncertainty.
- Why there is no real "rank #1" in AI answersThere is no fixed rank #1 in AI answers: lists vary by run, order may not mean preference, and results differ by assistant. Here is how to read position.
- How to read an AI answer like an analystRead an AI answer in five passes: who is named, in what role, in what order, what is cited, and what is claimed. Here is how to do it, step by step.
- How to compare answers across ChatGPT, Claude, Gemini and PerplexityAsk all four assistants the same question the same way, record each answer, then compare brands, roles and sources. Here is a fair method, step by step.
- How to spot when AI says something wrong about your businessAsk AI assistants factual questions about your business, check every claim against your own records, and log errors with the source. Here is the routine.
- AI visibility metrics that actually mean somethingRecommendation rate, citation rate, share of recommendations and cited-page presence mean something. A single mystery score does not. What to track and why.
03 Improving
- How to get recommended by ChatGPTChatGPT recommends brands it finds described clearly and positively on pages it trusts. Make your site readable, get listed on cited pages, and measure it.
- The kinds of pages AI assistants cite mostAssistants tend to cite roundups, comparisons, reviews, forum threads, documentation and vendor sites. Test your own questions to see which types matter.
- How to write a comparison page AI assistants trustA trusted comparison page is fair, specific and current. State your criteria, include real rivals, admit where you lose, and keep the facts checked and dated.
- How to get onto the roundup and review pages AI relies onFind the roundup and review pages AI assistants cite for your market, check whether you are listed, then earn a place with an honest pitch and a real product.
- AI crawlers and robots.txt: who can read your siteYour robots.txt decides which AI crawlers may read your site. Learn the difference between training, search and user-triggered crawlers, and how to check it.
- llms.txt explained: what it is and whether you need onellms.txt is a proposed plain-text file that points AI tools to your key pages. It is optional, and no assistant is known to need it. Here is when it helps.
- Structured data and AI answersStructured data (schema.org markup) helps machines read your pages, but no assistant promises to recommend you for it. What to mark up and how to test.
- Why your homepage confuses AI assistants, and how to fix itSlogans, vague claims and buried facts make a homepage hard to summarise. Say what you are, who it is for and what it costs in plain words near the top.
- Reviews, forums and community answers in AI recommendationsAssistants often draw on reviews, forum threads and community answers as independent voices. Take part openly and make it easy for real customers to speak.
- You launched something new: how long until AI notices?There is no fixed delay. Assistants that search the web can find a new page quickly once it is crawlable, but being recommended takes more than being found.
- How to fix outdated information in AI answersFind the source of the outdated claim, update your own pages, ask third parties to correct theirs, make current facts clear, then re-check over several weeks.
- What AI visibility means for agencies and their clientsFor agencies, AI visibility is a new reporting line: a fixed question set per client, regular measurement, honest uncertainty and clear next steps.
04 Categories
- What AI search visibility tools doAI search visibility tools ask ChatGPT, Claude, Gemini and Perplexity questions on your behalf, record the answers and report how often your brand appears.
- How to choose an AI search visibility toolChoose by what the tool measures, how it collects answers, whether numbers are traceable, and whether it fits your budget and reporting needs. A checklist.
- API or scraping: how AI visibility tools collect answersTools collect AI answers through official APIs or by automating consumer apps. Each has trade-offs for accuracy, stability and terms. Here is what to ask.
- Questions to ask before you buy an AI visibility toolAsk vendors how they collect answers, whether numbers trace to saved answers, how they handle variation, what the score is made of and what they cannot see.
- What is a link-in-bio tool?A link-in-bio tool gives you one link for your social profiles that opens a page of your links, content or products. Here is what it does and who it suits.
- Link-in-bio page or website: which do you need?Need a link-in-bio page or a website? Choose a link page for quick, mobile-first routing from social profiles, and a website for depth, search and ownership.
- How to choose a link-in-bio tool for a businessChoose a link-in-bio tool by the job it must do: capture leads, sell, report on clicks, and let your team manage it. A practical checklist for businesses.
- How link-in-bio pages show up in search and AI answersLink-in-bio pages rarely rank or get cited because they hold little text. Here is what search engines and AI assistants can read, and how to be found.
- What is live race tracking?Live race tracking shows participants' positions on a map while a race, ride or run is happening, so organisers, spectators and supporters can follow along.
- How to set up live tracking for a race or rideSet up live tracking in six steps: choose devices, test your course, brief participants, share one link, plan for lost signal, and prepare your supporters.
- Phones or dedicated trackers for live race trackingPhones are easy and free to use but need coverage and battery. Dedicated trackers cost more and work where coverage fails. Here is how to choose.
- What spectators need from live race trackingSpectators need a link that opens at once, a clear map, a findable runner, honest signs of freshness and a sensible finish estimate. Here is why.
05 Method
- Why every AI answer should be keptAI answers change and the web pages they cite change too. Keeping every full answer and cited page is what makes an AI visibility number checkable.
- Why AI visibility should be measured monthly, not dailyAI answers vary from one asking to the next, so daily numbers mostly show noise. A monthly edition, taken in one pass, shows real movement more clearly.
- How to build a benchmark set of questionsA good AI visibility benchmark is a fixed, public set of unbranded buyer questions, grouped by difficulty and intent, versioned so results stay comparable.
- Anchor questions: comparing AI answers across yearsAnchor questions are benchmark questions that never change, so the same wording can be asked in 2026 and 2030 and the answers compared. Here is how they work.
- Margins of error in AI measurement, explained simplyA margin of error shows how much an AI visibility number could move by chance. Here is how to read one, and why the question, not the answer, is the unit.
- Why an index of AI answers must never sell inclusionIf a brand can pay to be in an AI answers index, its numbers mean nothing. Here is why independence matters and how a trustworthy index protects it.
- Decision questions and informational questions: why the difference mattersSome questions ask an AI to choose products and others ask it to explain. Counting them together distorts a brand's numbers, so measurement keeps them apart.
- Does location change AI answers?Location can change AI answers, especially for local and regional questions. Here is how to test it and why measurement treats each market separately.
- When is a brand the same brand? Entity resolution in AI answersEntity resolution decides whether two names in AI answers are the same brand. Done badly it splits or merges brands and corrupts every figure built on it.
- One reader for four AIs: why consistent reading mattersComparing ChatGPT, Claude, Gemini and Perplexity is only fair if one reader applies one standard to all four answers. Here is how that works and its limits.
- How AI recommendations evolve, and what to watch forAI recommendations move as sources, models and brands change. Here is what to watch month to month, and how to tell a real shift from noise.
- A glossary of AI visibilityPlain-English definitions of AI visibility and AI answer measurement terms, from anchor questions and benchmarks to citations and recommendation rate.