Guides
How AI visibility actually works
Teach-first, data-led. Each guide cites the research a recommendation is built on. The product is the proof, not the pitch.
How to explain AI visibility to your CEO in one slide
One slide, four lines: where you are cited today per buyer question, the three fixes shipping this month, what you honestly expect to move and by when, and how you will know it worked. No pooled score, no competitor rank, no revenue forecast.
Check your own AI visibility from Claude or Cursor: the MCP workflow
You can run an AI-visibility scan on any domain from inside Claude Code, Claude Desktop, Cursor, Windsurf, or VS Code. Collimer ships an MCP server with exactly one tool, installed with a five-line config block. Below: the setup, three prompts that work verbatim, and an unedited transcript of a real session, including our own unflattering score.
What GEO costs in 2026: DIY, tool, agency, or sprint, and how to choose
GEO costs time before it costs money. The honest ranges: $0 and 2 to 3 hours a week for DIY, $20 to $800 a month for a tracker, $1,500 to $25,000 a month for an agency retainer, and a fixed-scope sprint that ends cleanly in between. The route that fits depends on one question: can you ship a site change yourself this week?
AI-visibility tools with hidden pricing
Some AI-visibility and GEO tools don't publish dollar figures at all: Conductor and BrightEdge are sales-gated with no public tiers, and Peec's prices are JS-rendered and invisible in the page's raw HTML. That last case is the same failure Collimer warns about elsewhere: content only AI crawlers can't read is content AI crawlers can't cite.
GPTBot, ClaudeBot, PerplexityBot: which AI crawlers to allow, and what each one does
Every major AI vendor runs separate bots for training, for its search index, and for live user-triggered fetches. Blocking the wrong one makes you invisible in AI answers without actually opting out of anything. A verified, bot-by-bot reference, as of August 2026.
Scrunch AI pricing (2026)
Scrunch AI costs $250/mo for Starter (billed annually; $300/mo month-to-month) and $417/mo for Growth (annually; $500/mo month-to-month), verified against Scrunch's own pricing page in August 2026. Scrunch's Starter price lands on the exact same number as Collimer's Team plan, priced for a different product.
AI search is eating your organic traffic: what to do this quarter
Rankings held, sessions fell: the clicks moved inside the AI answer. A calm one-quarter plan for a marketing lead with no analyst: diagnose branded vs problem-led visibility, build the five pages that answer buyer questions, get corroborated off-site, and measure per engine.
Profound pricing (2026)
Profound costs $99/mo for Starter (ChatGPT only, 50 prompts), $399/mo for Growth (3 engines, 100 prompts), and custom Enterprise pricing above that, verified against Profound's own pricing page in August 2026. It is a tracker with content-generation ("Agents") credits bolted on, built for teams with an analyst who will act on a trend.
GEO tool pricing compared (2026)
GEO and AI-visibility tool pricing runs from about $20 a month to $800 a month for self-serve tiers, verified against every listed vendor's own pricing page in August 2026. The price gap tracks a real product gap: most of the market sells a number to watch, a small minority sells a ranked, shipped, re-verified fix.
How to get your B2B SaaS cited by ChatGPT: a 90-day plan
Getting cited is a sequencing problem: be describable, then retrievable, then corroborated, then verify with a frozen panel. The phases in order, the hours each takes for a team of one, and what honestly does not move in 90 days.
How to tell whether a GEO fix actually worked
A single re-check cannot distinguish a real visibility change from run-to-run noise. The honest method: freeze the prompt panel, re-probe with a confidence interval, and report one of five outcomes, including the null.
How fast can AI visibility change?
In one measured case, own-domain AI citation rate rose from 1.25% to 5.8% within two weeks of a targeted fix, confirmed by two independent tools. Why AI visibility can move in weeks when the work targets what assistants actually read, and what that speed does and does not prove.
A broken Markdown table is getting cited by AI engines
Mass-producing unvalidated AI content can win AI citations today, but the citations it wins rarely land near a real buying decision, and the loophole that makes it work is already closing. The clearest proof: a real comparison page whose feature table is raw, broken Markdown, visibly wrong to a human, and still gets cited by AI answer engines.
Your AI visibility score is one number. Your product isn't.
A pooled AI-visibility score describes your brand's average, not your product lines. One benchmark scan found a company dominating its core categories while it was nearly invisible in three newer ones under the same brand, a gap a single score fully masked. The evidence is one case, not a shipped detector, so read it as worth checking rather than a rule.
Your GitHub README is a GEO surface
One open-source developer-tools company rewrote its GitHub README and standardized metadata across 16 repos, about two person-days of work. Its own-domain AI citation rate rose from 1.25% to 5.8%, a 4.7x lift, in two weeks, confirmed independently by two separate measurement tools. No paid placement, no site redesign.
An apostrophe cost this brand 43 visibility points
The same brand, scanned twice minutes apart with only its brand-alias spelling changed, scored 5 out of 100 and then 48 out of 100. Nothing about how AI engines regarded the business changed between the two runs. What changed was whether the scan's matcher recognized the name the way the world actually writes it.
What's new: your report now shows how AI positions you
Between July 21 and July 24, 2026, six merged pull requests added a "How AI positions you" section to every Collimer report: a one-line verdict plus up to four insight cards, each linked straight to its fix in the roadmap.
A good AI visibility score can still mean zero new customers
A pooled AI-visibility score can be built almost entirely from branded queries, the ones from buyers who already know your name. Split the same score by query category and a different number often shows up: near-zero visibility on the problem-led queries where new demand actually starts, with a competitor cited there instead.
You're mentioned. Are you described correctly?
A mention count only proves an AI engine said your name. It does not prove the sentence around your name described the right company. Two distinct identity failures hide behind a healthy-looking visibility score, and they need different fixes.
What's new: Collimer now reads local and service businesses, not just SaaS sites
Collimer's measurement pipeline was originally tuned against one design partner, a local gym. The week of July 13, 2026, four fixes generalized it: service-generic probe seeds, a brand-alias fix, a local/service fallback, and a real-location field.
Why AI says it doesn't know your brand yet (and what actually changes that)
A 0/100 AI-visibility score almost never means the model looked at your brand and ranked it last. It means the model has no memorized knowledge of the brand at all, a distinct failure mode that hits small and local businesses hardest, and one a grounded search path can often recover from even when the model's memory can't.
GEO isn't a SaaS-only game: you're found for the problem, not your industry
AI-visibility work looks SaaS-shaped because that's who talks about it loudest, but the mechanism routes on the problem a searcher is trying to solve, not the vertical a business sits in. A local-fitness scan and a services agency's own numbers show the same pattern a Series B SaaS company sees.
If one AI engine goes dark mid-scan, is your visibility score still trustworthy?
A pooled AI-visibility score built from five engines is only as trustworthy as its weakest disclosure. When one engine returns zero successful probes during a scan, silently pooling that silence into the survivors' rate produces a number that looks whole but was measured on a partial panel. The fix is disclosure, not a new formula.
How to measure AI visibility properly (and why one-shot checks lie)
A single AI-visibility scan is a snapshot, not a measurement. Comparing two scans only means something if the engine set, the prompts, and the repetition count are all held fixed between them. Here's why, and what a trustworthy visibility number actually requires.
Can you buy AI visibility? What paid placement actually gets you
Every major AI engine that runs ads keeps the ad architecturally separate from the organic answer. Paid placement buys adjacency to an AI answer, not a mention inside it. Here's the evidence, and where the budget actually belongs instead.
Why your Copilot visibility number is an estimate, not a measurement
Microsoft Copilot has no public API, so any Copilot AI-visibility number, ours included, is a proxy synthesized from web-search snippets, not a direct reading. What that means for how much weight to put on that one number.
How five AI engines each decide what to cite
ChatGPT, Perplexity, Gemini, Grok, and Claude don't work the same way. They differ in whether they ground at all, how they return citations, and what content they reach for. A strategy that works on one may actively underperform on another.
Schema markup for AI visibility: what works and what doesn't
FAQPage and HowTo schema are confirmed non-levers for AI citations. SoftwareApplication schema matters for B2B SaaS on Gemini and Bing Copilot. And blocking Google-Extended doesn't do what most teams think it does.
Three content signals that lift AI citation rates by 30 to 41 percent
A 2024 Princeton study tested nine GEO optimizations across 10,000 queries. Three of them, adding statistics, quotations, and cited sources, each lifted generative-engine citation rates by 30 to 41 percent. Keyword stuffing underperformed.
What AI crawlers actually read on your site (and what they skip)
GPTBot, ClaudeBot, and PerplexityBot execute zero JavaScript. Content that only exists after client-side rendering is invisible to them. Making content readable by agents is a different discipline than making it readable by browsers.
What is GEO? How generative engine optimization differs from SEO
GEO and SEO optimize for different things: SEO targets crawl, index, and rank; GEO targets retrieval and citation inside AI-generated answers. They share some tactics, but the mechanism, and the measurement, is different.
Why your AI visibility is probably lower than you think
Most brands check one AI engine, see their name, and move on. The problem: a positive result on ChatGPT tells you almost nothing about Perplexity, and neither reading tells you whether you're being found when buyers are still figuring out what they need.
Third-party sources win evaluative queries
When buyers ask an assistant "what's the best tool for X", the engines cite other people's pages far more than your own. Why your comparison page isn't enough, and what is.
Brand mentions beat backlinks for AI visibility
Unlinked mentions of your brand across the web track AI-search visibility far more closely than backlinks do. The evidence, the caveat, and what to do about it.
Front-load the answer: the single highest-leverage GEO fix
Pages that state the answer in the first paragraph get cited far more often by AI assistants than pages that build up to it. The evidence, and how to apply it.
Reviewed