Words we use
Every term in this report, in plain words, with where to see it in action.
- AI engine
- An AI assistant people ask questions: ChatGPT, Claude, Gemini, Perplexity. We asked each the same questions, the way a person would. Read their answers
- Question
- One thing a real person might ask. There are 2,290 in all, and none names an Atlanta organization, so no one gets a head start from being in the question. Read every question and answer
- Answer
- One question asked of one engine. It's the unit we count: a figure like “named in 40” means 40 answers. Read the answers
- The 340 questions everyone faced equally: 170 about Atlanta's ecosystem and 170 about US startup cities. The only numbers we set side by side come from here. How the questions are built
- Own study
- A study of one organization, from its own buyers' questions. There are 13. They show how an organization does on its own ground, so they're never compared with each other. Open an own study
- Named
- An answer mentions the organization or city at all, anywhere in the text. Every organization, by role
- Picked (recommended)
- An answer singles it out as a top recommendation, not just a mention. This is our coding of the answer; quotes are always the engine's own words. Every organization, by role
- Cited (source)
- A web page the engine listed as a source for its answer. A citation shows the engine read the page, not that the page wrote a particular sentence. Where AI gets its information
- From memory
- An answer that cites no sources at all. The engine answered from what it already knew, which tends to favor well-known names. The memory lesson
- Share of voice
- An organization's slice of all the mentions of the 63 organizations we track. If an answer names three, each gets one. Every organization, by role
- Win rate
- Of the answers that name an organization, how many pick it. It shows whether being mentioned turns into being chosen. Every organization, by role
- Top five
- When an answer gives a ranked list, whether a city lands in its first five places. The finding
- Head-to-head
- A question that puts Atlanta directly against one other city, such as “Atlanta or Austin?” Read the head-to-heads
- Cluster
- A group of questions behind the same kind of decision, like choosing a program or raising money. Patterns are checked across a whole cluster, not one answer. See the clusters
- Buyer's job
- In an own study, what the person asking is trying to do: name a problem, make a shortlist, compare, or check a choice. How own studies are built
- Home field
- The gap between how an organization does on its own buyers' questions and on the market's shared ones. Many look stronger at home. The home-field lesson
- Move
- A small, specific step an organization can take, drafted from what the engines say today: who does it, how long it takes, and words to paste. There are 204. Find your organization
- Re-measure
- Asking the same questions again, the week of December 7, 2026 (date not final), to see what changed. The tracker
Published by PursueATL. Research by Resonate Labs. Data collected September 2026.