SEO and AEO glossary
36 terms you meet when buying an AI SEO tool. Each one gets a plain definition and the thing people most often get wrong about it, because the second part is usually what costs money.
Written by Ron Kloth, founder of RankJudge.
AI search and answer engines
The vocabulary that did not exist three years ago, and the part of search most tools are now sold against.
AEO (Answer Engine Optimization)
Optimising so AI assistants like ChatGPT, Claude and Perplexity name your site when they answer a question. The work is factual accuracy, sources an engine can check, and structure it can parse.
What people get wrong: It overlaps with Google ranking less than people assume. In a dataset of 100 prompts run through ChatGPT, Claude and Gemini on 7 October 2026, only 8%, 3% and 14% of the pages each engine cited were in Google’s US top ten for the search it had just run. Ranking first does not get you cited. What does carry over is factual accuracy and structure an engine can parse, and the tools selling "AEO" as a new category mostly sell tracking rather than optimisation.
Compare two AI visibility trackers →GEO (Generative Engine Optimization)
The same idea as AEO under a different name, covering generative systems broadly: ChatGPT, Google AI Overviews, Gemini, Copilot.
What people get wrong: GEO and AEO are used interchangeably by most vendors, and the ones insisting on a distinction usually have a product that maps to their preferred term. Do not pay extra for a label.
AI Overview
Google’s AI-written summary at the top of a results page. It answers the question inline, so the user often never clicks a link.
What people get wrong: Blocking Google-Extended does not remove you from AI Overviews. That control only governs training for Gemini; AI Overviews are built from the ordinary search index and follow Googlebot.
Check which crawlers your site allows →Citation (in AI answers)
When an assistant names your site as a source. It is the AI equivalent of a ranking, and the thing AI visibility tools exist to count.
What people get wrong: A citation and a click are different events, and most citations produce no visit at all. Analytics can show you referral traffic from an assistant; it cannot show you the times you were named and nobody clicked.
Tools that track AI citations →Answer engine
Any system that responds with an answer instead of a list of links: ChatGPT, Perplexity, Claude, Google AI Mode.
What people get wrong: They do not all get their information the same way. Several lean on an existing search index rather than their own crawl, which is why being absent from Bing can quietly remove you from an assistant that never mentions Bing.
llms.txt
A proposed file offering AI systems a clean summary of a site and links to its important pages.
What people get wrong: It is a proposal, not a standard, and Google has stated it does not use it. Having one is harmless; paying anyone to optimise it is not a purchase we can justify today.
See whether your site has one →Open Knowledge Format (OKF)
A Google Cloud specification for packaging knowledge as a folder of markdown files, each with a YAML header recording where the content came from, who last verified it and when it goes stale.
What people get wrong: It is a real published spec rather than a proposal, which is more than llms.txt can claim, but it is not an SEO file. The specification never mentions crawlers or search, lists "prescribing storage, serving, or query infrastructure" among its non-goals, and defines no web address to serve it from. Nothing crawls for it today, so anyone selling OKF as a way to get cited is selling you something the spec does not do.
Read ours →AI crawler
An automated visitor sent by an AI company. Some fetch your page to answer a live question; others collect content to train future models.
What people get wrong: Treating the two as one thing is the expensive mistake. Blocking a training crawler is a defensible business decision. Blocking an answering crawler makes you invisible to that assistant, and most sites that do it never meant to.
Free check: which bots can read your site →Share of voice (AI)
How often your brand appears in AI answers for a set of prompts, against competitors. The headline metric on most AI visibility dashboards.
What people get wrong: It is only as meaningful as the prompt list behind it, and vendors rarely publish how prompts are sampled or how often they refresh. Ask before you buy, because that mechanic decides whether the number means anything.
Hallucinated citation
When an assistant attributes a claim to a source that never said it, or invents the source outright.
What people get wrong: This cuts both ways. It can credit you for something you never published, which inflates a tracking dashboard, and it can attribute your work to a competitor.
Finding what to write about
The research end. This is what most of the cheaper tools in the category actually sell.
Keyword research
Finding the terms people search that relate to your business, and judging which of them you could realistically rank for.
What people get wrong: The trap is chasing volume. A term with 40,000 searches that you cannot rank for is worth less than one with 200 that you can, and the whole value of a keyword tool is telling those apart.
Two budget keyword tools compared →Long-tail keyword
A longer, more specific phrase with lower volume and clearer intent. "Best AI SEO tool for agencies" rather than "SEO tool".
What people get wrong: Long-tail does not automatically mean easy. Plenty of long phrases are dominated by vendors with large domains, and specificity alone is not weakness.
A long-tail specialist compared →Keyword difficulty
A score estimating how hard a term is to rank for, usually built from the strength of the pages currently ranking.
What people get wrong: Every vendor calculates it differently and none of them publishes the formula. Two tools can score the same keyword 20 points apart, so it is a rough sort, never a decision.
Search intent
What somebody is actually trying to do when they type a query: learn, compare, or buy.
What people get wrong: It is read off the results page, not guessed. If a query returns nine comparison pages, Google has already told you what it thinks the intent is, and a page that answers a different one will not rank however good it is.
SERP (search engine results page)
The page of results Google returns: organic links, ads, featured snippets and now AI Overviews.
What people get wrong: The organic slots have moved down the page. Counting a position without looking at what sits above it overstates what that position is worth.
Featured snippet
The boxed answer at the top of some results pages, lifted from a page Google considers the best direct answer.
What people get wrong: Winning one does not reliably win you the click. The answer is on the results page already, which is the same trade AI Overviews make on a larger scale.
Cannibalisation
When several of your own pages compete for the same term, so none of them establishes itself.
What people get wrong: It is usually caused by publishing volume without a plan for what each page is for, which makes it a particular risk for anyone buying a tool to publish faster.
Generating volume, and the cost of it →Measuring and ranking
The numbers tools report, and how much weight each one actually carries.
Rank tracking
Monitoring your position for chosen keywords over time.
What people get wrong: Results are personalised and vary by location and device, so a tracker reports one sample of a moving target. Trends over weeks are useful; a single day’s position is close to noise.
Tools by what you need them for →CTR (click-through rate)
The share of people who see your result and click it.
What people get wrong: A falling CTR is not automatically your fault. When an AI Overview or a snippet answers the query inline, clicks drop at an unchanged ranking position.
What AI answers did to click-through rates →Domain authority
A third-party score, usually 1–100, estimating how likely a site is to rank. Moz popularised it; other vendors have their own version.
What people get wrong: Google does not use it. It is one company’s model of Google, useful for comparing two sites on the same scale and meaningless as a target in itself.
Ahrefs and Moz Pro compared →Backlink
A link from another site to yours. Still one of the strongest signals in search.
What people get wrong: Source matters more than count. A handful of links from sites in your field beats hundreds from a network selling them, and the paid schemes that promise volume are the ones Google’s guidelines name.
Two tools that build links →Crawl budget
How much attention a search engine spends on your site: how often it visits and how many pages it fetches.
What people get wrong: Small sites rarely have a crawl budget problem, but new ones have a crawl *interest* problem. Pages sitting unindexed usually mean the engine has not decided the site is worth the visit yet, which more pages will not fix.
Indexation
Whether a page is actually in a search engine’s index and therefore able to appear at all.
What people get wrong: Published is not indexed. "Discovered - currently not indexed" in Search Console means Google knows the URL and has chosen not to crawl it, which is a judgement about the site rather than a technical fault.
Benchmark
A published figure from a named organisation, used to interpret whether a result is good.
What people get wrong: Check its date and its source before repeating it. Figures in this field go stale within months and get recirculated for years. The most quoted AI market-share number we found was wrong by more than twenty points.
Our benchmarks, each dated →Buying the tools
The commercial vocabulary. This is where comparison pages are least honest, so it is where the definitions matter most.
Content optimiser
A tool that grades a draft against the pages currently ranking and tells you what to change. It edits; it does not generate at volume.
What people get wrong: An optimiser and an AI writer are different products and are constantly compared as if they were alternatives. One fills a blank page, the other marks it. Buy for the job you are short of.
A writer and an optimiser compared →Annual-billed-monthly
The price you pay per month when committing to a full year up front. Almost always the number shown by default on a pricing page.
What people get wrong: It is not the monthly price, and quoting it as one is the single most common error in this category. Switch the toggle both ways before believing any comparison, including ours.
How we check a price →Seat
One user login. Plans include a set number and charge for more.
What people get wrong: Seats reorder the cheapest-tool question faster than features do. A tool $20 cheaper that charges per extra user is the expensive one the moment a second person needs access.
Where seats decide the answer →Credits
A metered allowance of analyses, generations, or queries, consumed as you work.
What people get wrong: Credits are not comparable between vendors. One tool’s credit is an article, another’s is a single page analysis, so comparing two plans on credit count alone compares nothing.
Overage
What happens after you exhaust a plan’s allowance: extra charges, or a stop until the month resets.
What people get wrong: Most pricing pages do not state it. That absence belongs in your decision. A plan whose behaviour at the limit is undocumented is a plan with an unknown bill attached.
Annual lock-in
A contract committing you to twelve months, usually in exchange for a lower monthly rate.
What people get wrong: The discount is real and so is the risk. The tools people abandon are abandoned in the first six weeks, which is exactly the period an annual plan removes your exit from.
White-label
Reports and dashboards carrying your branding instead of the vendor’s, so an agency can put them in front of clients.
What people get wrong: Availability is usually tier-gated and rarely on the entry plan. If reselling is the reason you are buying, confirm which tier carries it before comparing prices.
Two platforms agencies resell →Schema markup
Structured code that tells search engines what a page is: a product, an article, a person, a set of questions.
What people get wrong: It makes a page easier to understand; it does not make it rank. It is also where overclaiming gets caught, since review markup demands a rating and inventing one to win a rich result is the kind of thing that gets noticed.
Affiliate link
A tracked link that pays a commission if a reader buys through it, at no extra cost to them.
What people get wrong: It is disclosure that makes one acceptable, not the absence of one. RankJudge uses affiliate links, says so under every button, and publishes the full arrangement, including that we run no testing of the tools we write about.
How this site makes money →API access
Permission to pull a tool’s data into your own systems or reports.
What people get wrong: Usually a top-tier feature, and usually metered separately from the plan’s headline allowance. Two separate limits, not one.
BYOK (bring your own key)
A pricing model where the subscription buys the software and you pay the AI provider separately, using an API key you create yourself at OpenAI, Anthropic, Google or similar. The tool sends your prompts; the model bill arrives from someone else.
What people get wrong: It is not a lower price, it is a different person paying. A BYOK plan and an all-inclusive plan cannot be compared on their subscription figures, because one of the two numbers is missing from the page you are reading it on. Work out the per-article or per-task cost of the inclusive plan from its own cap, and you have the break-even your own API spend has to beat.
A BYOK tool against an all-inclusive one →Free trial vs free tier
A trial gives temporary access, often to an expensive plan. A tier is permanently free with hard limits.
What people get wrong: They suit different buyers and comparison pages treat them as the same offer. A trial tells you whether the top plan is worth it; a tier tells you whether the product fits your workflow at all.
Two very different free offers →Know the words. Now pick the tool.
Definitions only matter up to the point of a decision. We name one pick and one runner-up for nine kinds of business, with every price read off the vendor's own page and dated. If AI search is new to you, ourguides and resources are the place to start instead.
Find your tool