THE WORDS, DEFINED ONCE.

AEO is the work of getting a business named and cited inside the answer an AI assistant generates, as distinct from ranking in a list of links. Its inputs are one clean, disambiguated entity; valid structured data that agrees with the visible page; content written so a direct answer can be lifted whole; llms.txt; explicit permission for AI crawlers; and third-party citations the assistant already trusts. Unconventional Group sells it as a one-time engineering pass at $1,500–$2,000, separately from SEO.

Why it matters to a business owner: when a buyer asks an assistant for "a good X in Edmonton" and gets three names back, there is no page two.

GEO is another name for AEO: making a business visible to generative AI systems such as ChatGPT, Claude, Perplexity, Gemini and Google's AI results. The two terms are used interchangeably in practice, with GEO slightly more common in academic writing and AEO slightly more common among practitioners. Some people also say "AIO" or "AI visibility". They describe the same work.

Why it matters to a business owner: if a vendor sells you both AEO and GEO as separate line items, ask what is different.

llms.txt is a plain-text file at the root of a website that gives AI systems a curated, first-party description of what the business is, what it sells, and what it does and does not claim. It is a proposed convention rather than a ratified standard, and evidence for how much weight assistants give it is still thin. It is cheap to publish and, written honestly, it is the lowest-cost guardrail against being described inaccurately. A wrong or stale one actively hurts, which is why Unconventional Group's Odin auditor checks that it agrees with the live site.

Why it matters to a business owner: it is the one place you get to state your own facts in the words you want quoted.

llms-full.txt is the long-form companion to llms.txt: a complete, machine-readable reference an AI system can read in one fetch instead of crawling the whole site. Where llms.txt is a summary with links, llms-full.txt carries the substance: services and prices, terms and limits of what is claimed, service areas, disambiguation from similarly named companies, and direct answers to common questions. It must be kept in sync with the site or it becomes a liability.

Why it matters to a business owner: an assistant that reads one accurate document is less likely to invent details than one assembling your story from fragments.

A Markdown twin is a plain-text version of a web page, served alongside the HTML, so AI systems can read the content without parsing layout, scripts and navigation. It is typically exposed at a predictable address (for example the page URL followed by .md or /md) and generated from the same source as the page so the two cannot drift. Unconventional Group's flagship AEO case study, the 126-page AltaPro AI site, ships a Markdown twin of every page.

Why it matters to a business owner: it removes the most common reason an assistant misreads a page, which is that the page was built for eyes rather than parsers.

schema.org is the shared vocabulary that websites use to describe their content to machines: Organization, LocalBusiness, Service, Offer, FAQPage and hundreds of other types. It is maintained jointly by the major search engines and is what "structured data" and "schema markup" refer to. The vocabulary defines what can be said; JSON-LD is the usual format for saying it.

Why it matters to a business owner: it is how you tell a machine "this is a plumbing company in Sherwood Park that charges from $X" without hoping it infers that from your homepage.

JSON-LD is the format, embedded in a page's HTML, in which schema.org structured data is most commonly written. It sits in a <script type="application/ld+json"> block, separate from the visible content, and can describe several related nodes in one graph: the organization, the website, the page, a service and its offer. Google recommends it over the older microdata and RDFa formats. A price in the JSON-LD must match the price on the page, or the markup is worse than none.

Why it matters to a business owner: it is the plumbing behind rich results in search and behind an assistant being able to quote your price correctly.

FAQPage is the schema.org type for a page that contains a list of questions, each with a single accepted answer, marked up so machines can read the pairs directly. The markup must match what is visible: a page with FAQ markup and no visible FAQ is a policy violation. One FAQPage node per URL is correct; several on the same page is worse than one. Google has narrowed which sites get FAQ rich results, but the markup remains a clean way to hand an assistant an extractable answer.

Why it matters to a business owner: it is the most direct way to put a quotable answer to "how much does X cost" in front of a machine.

An entity is a specific, uniquely identifiable thing in the world, such as a particular company, person or place, as opposed to a word that could refer to several. Search engines and AI systems reason about entities, not keywords: "Unconventional Group" is a string, but the Edmonton company at unconventionalgroup.ca, established 2023, founded by Haruun Ali, is an entity. A site should have exactly one node describing its business entity, with a stable @id, and every other node should reference it.

Why it matters to a business owner: a machine cannot recommend you until it is sure which "you" it is talking about.

Entity disambiguation is the explicit work of telling machines which business you are when your name could be confused with another. Common words in company names ("Unconventional", "Summit", "Premier") make this necessary. It is done with a disambiguatingDescription in structured data, a stated location and founding year, consistent NAP, external identifiers, and a plain-language note in llms.txt saying what you are and what you are not. Unconventional Group's own llms.txt notes that unrelated agencies use the word "Unconventional" and identifies itself by domain and city.

Why it matters to a business owner: without it, an assistant may describe a competitor's reviews and prices as yours.

sameAs is a schema.org property that links an entity on your site to the same entity's profiles elsewhere, such as its Google Business Profile, LinkedIn page, Wikidata item or an affiliated company's site. It tells a machine that these separate URLs all refer to one thing. The strongest signal is when the relationship is stated from both sides: your site points at the profile, and the profile points back at your site.

Why it matters to a business owner: it turns scattered profiles into one corroborated identity.

A knowledge graph is a database of entities and the relationships between them, which search engines and AI systems consult to understand who and what a query refers to. Google maintains a private one; Wikidata is the public, editable one that many systems draw on. A business qualifies for a Wikidata item only when it meets notability rules, so for most small businesses the groundwork is consistent structured data and third-party citations rather than a Wikidata entry.

Why it matters to a business owner: being in the graph is how you go from "a plumber in Leduc" to a named, recommended company.

An AI crawler is a bot operated by an AI company to fetch web pages, either to train models or to retrieve sources for live answers, identified by a user-agent string in robots.txt. The ones Unconventional Group names and allows on its own site are GPTBot, OAI-SearchBot and ChatGPT-User (OpenAI), ClaudeBot (Anthropic), PerplexityBot and Perplexity-User, Google-Extended, Applebot-Extended, Bytespider, meta-externalagent, cohere-ai and CCBot. Some fetch for training and some for answering; blocking the answering ones removes you from live answers.

Why it matters to a business owner: a platform default or a security plugin may be blocking these without anyone having decided to.

Citation rate is the share of a fixed set of AI-assistant prompts whose answers cite a specific business's own domain as a source. It is stricter than mention rate: an assistant can name a business from a directory listing without ever citing the business's site. A rising citation rate means the assistant is reading and trusting first-party pages. Because assistants are nondeterministic, each prompt is run several times per period and the rate is reported, not a single yes or no.

Why it matters to a business owner: a citation sends the buyer to your site; a bare mention may send them to a directory.

Mention rate is the share of a fixed set of AI-assistant prompts whose answers name a specific business at all, whether or not they cite its website. It is the broadest measure of AI visibility and the first one to move when third-party citations and entity clarity improve. It is reported per assistant, because ChatGPT, Claude, Perplexity and Gemini retrieve differently and rarely agree.

Why it matters to a business owner: it answers the plain question "does the assistant know I exist".

Share of voice is the proportion of business mentions across a prompt set that belong to one business rather than to its competitors. If a set of prompts produces 100 named businesses and 20 of them are yours, your share of voice is 20 percent. It is the metric that puts mention rate in context: being mentioned in half of prompts matters less if every competitor is mentioned in all of them.

Why it matters to a business owner: it tells you who the assistants are recommending instead of you.

A prompt set is the fixed list of buyer-intent questions used to measure AI visibility, held constant so that results are comparable from one period to the next. Unconventional Group's tracking product defines 20 to 40 prompts per client, written the way a real buyer types ("who is the best commercial electrician in Edmonton"), not as keyword strings. The set is fixed because a metric is only comparable over time if the question does not change.

Why it matters to a business owner: it turns "are we visible to AI" from a feeling into a number you can re-run.

AI visibility tracking is the recurring measurement of whether and how AI assistants mention, cite and rank a business, scored against a baseline. Unconventional Group's product is $200 a month with a three-month minimum: a fixed prompt set is re-run on a schedule across the target assistants, every raw response is stored, and each run is scored for mention rate, citation rate, average position and share of voice, rolled into a 0–100 score with a published formula. The three-month minimum is methodological: one month of data cannot separate a trend from assistant noise.

Why it matters to a business owner: it is the instrument that tells you whether the AEO work, or anything else, changed what the assistants say.

Answer-shaped content is writing in which the first sentence of a section states the direct answer to the question the heading asks, with qualification and detail following it. It names the entity ("Unconventional Group's website builds start at $3,000") rather than using a pronoun, and it states concrete numbers where numbers exist. A paragraph that builds toward its point cannot be lifted whole and attributed; a paragraph that leads with it can.

Why it matters to a business owner: assistants preferentially quote sources that answer outright, and "it depends, contact us" is unquotable.

Nondeterminism means an AI assistant can give a different answer to the same question asked twice, because its retrieval and generation both vary from run to run. This is why a single check ("I asked ChatGPT and it named us") proves little, why prompts are run multiple times and reported as rates, and why no vendor can guarantee an assistant will recommend a business. The honest claim is a measured change in rate over months, not a promised placement.

Why it matters to a business owner: it is the reason to distrust any guarantee about AI recommendations, including from suppliers who seem confident.

Cold email infrastructure is the set of domains, inboxes, authentication records, warming schedules and monitoring that determines whether outbound email is delivered to an inbox at all. Unconventional Group's lead generation service is built on it: separate sending domains, provisioned inboxes, SPF/DKIM/DMARC, warmup, verified lists, sequencing, deliverability monitoring and reply routing into the client's CRM. It is scoped per engagement because cost tracks inbox volume and list size, and it is not advertising.

Why it matters to a business owner: most failed outbound fails here, invisibly, before anyone reads the copy.

A sending domain is a separate domain, related to but distinct from a company's primary domain, provisioned solely for outbound email so the primary domain's reputation is never at risk. If example.ca handles invoices and customer email, outbound goes from something like example-team.com or tryexample.ca. Each sending domain carries a small number of inboxes kept within safe daily volumes.

Why it matters to a business owner: a cold campaign that damages the domain your invoices come from is a campaign that cost more than it earned.

Domain warming is the deliberate, gradual ramp of sending volume from a new domain or inbox so that mailbox providers learn to trust it before it sends at scale. A new inbox that sends hundreds of messages on day one is filtered to spam and tends to stay there. Warmup runs for weeks before a single prospect is contacted, which is the main reason a properly built outbound system takes weeks to reach full volume.

Why it matters to a business owner: it is why "we can start sending Monday" is a warning sign, not a selling point.

Deliverability is the share of sent email that reaches the recipient's inbox rather than the spam folder or being rejected outright. It depends on authentication, domain and IP reputation, bounce and complaint rates, list quality, sending volume and content. It is not a setup task that completes but a condition maintained through placement testing, blacklist monitoring and resting inboxes when the numbers say so.

Why it matters to a business owner: reply rate is capped by deliverability; no copy can fix a message nobody sees.

SPF is a DNS record that lists which mail servers are allowed to send email on behalf of a domain. A receiving server checks the sending server against the list; a message from an unlisted server fails the check. Each domain gets one SPF record, and it has a lookup limit that is easy to exceed when several services (email provider, CRM, marketing tool) are all added to it.

Why it matters to a business owner: without it, anyone can send mail claiming to be you, and providers treat your real mail with suspicion.

DKIM is a cryptographic signature added to each outgoing email that lets the receiving server verify the message was not altered and was authorized by the domain. The public key is published in DNS; the sending service signs with the private key. It has to be set up per sending service, and a missing or broken DKIM record is a common reason legitimate mail lands in spam.

Why it matters to a business owner: major mailbox providers now expect it from bulk senders, and its absence is enough to be filtered.

DMARC is a DNS policy that tells receiving servers what to do with email that fails SPF or DKIM checks, and where to send reports about it. Policies range from none (monitor only) through quarantine to reject. It also requires the visible From address to align with the domain that passed authentication. Large mailbox providers require a DMARC record from anyone sending at volume.

Why it matters to a business owner: it closes the loop on authentication and is now a condition of getting bulk mail delivered at all.

List building is turning a defined ideal customer profile into a real list of named contacts; enrichment adds the fields a sequence needs; verification confirms each address exists before anything is sent. Firmographic filters (industry, size, region) and role targeting narrow the list; enrichment supplies the specifics that let a message say something true about the recipient. Purchased lists resold to competitors and scraped lists sent without verification are how domains get burned.

Why it matters to a business owner: outbound rewards a specific, verified list and punishes a vague or bought one.

A sequence is a planned series of outbound messages sent to a contact over days or weeks, with rules that stop it the moment the contact replies. Good sequences have real variation between steps, are written to be read by a person rather than assembled from a template, keep per-inbox volume inside safe limits, and carry accurate sender identification and a working unsubscribe in every message.

Why it matters to a business owner: the follow-up, not the first message, is where most replies come from.

CASL is the federal law, in force since 2014, that governs commercial electronic messages sent to or from Canada, requiring consent, sender identification and a working unsubscribe mechanism. Consent can be express or implied; for business-to-business email, implied consent commonly rests on the recipient having conspicuously published their business address without a no-solicitation notice, where the message is relevant to their role. Every message must identify the sender, give valid contact information, and offer an unsubscribe that is honoured promptly. This is not legal advice; counsel should review any outbound setup.

Why it matters to a business owner: the penalties are serious, the requirements are specific, and "everyone does it" is not a defence.