Knowledge base

If your brand isn't in the generative answer, the user often never reaches the click.

If your brand isn't in the AI response to a commercial query ("which one to pick", "where's it cheaper", "what to buy"), you lose the sale before the click — the decision happens inside the model.

AI providers behave differently: a brand can be prominent in Alice and invisible in ChatGPT or Perplexity. Measure GEO across several models, not one.

Brands that land in AI answers now get baked into model memory — the first-mover advantage is real.

Inside, the end-to-end FinPilot teaching case: why a brand with a decent site still drops out of AI answers, and which signals to strengthen first.

A one-off audit is a snapshot — a repeatable cycle is control. That's what separates sustained AI-visibility growth from occasional spikes.

Start with prompts your customers would actually use to find you. A broad query like "freelance marketplace" creates noise in the competitor list; a narrow one ("where to find a freelancer for a WordPress online store") surfaces direct competitors and clean signals you can act on.

The first audit is built from a single run — it is a starting point, not a verdict. Look first at the competitor list and the source map: they show which platforms you need to appear on for the models to start naming you more often.

Import prompts in bulk one topic at a time and assign the group right away: metrics are calculated per cluster, and they only mean something when the queries inside are homogeneous. Keep the list larger than your limit — inactive prompts cost nothing — and activate the directions you are actually fighting for visibility in right now.

Start with the clusters that have the fewest prompts — that is where extending coverage moves visibility the most. And dismiss what does not fit right away: dismissed queries never resurface, so the recommendation list gets sharper over time.

Connect the bot right after your first run: the report arrives on its own, so you spot a visibility drop during the week instead of the next time you happen to open the dashboard.

If your team works across several brands in different workspaces, invite people surgically — only into the brands where they actually contribute. Credits and the subscription are scoped to the brand, not the user, so adding a teammate doesn't increase your bill.

Don't rush to optimize right after launch — give the system at least 3–5 days to gather data. The more observations, the more reliably real patterns and weak spots surface instead of random noise in one or two AI answers.

Start your analysis with dynamics, not absolute values. A 16% SoV is excellent if it was 5% a month ago and alarming if it was 25%. Always look at the change vs the previous period and at the "Winners & Losers" block: it shows which sources started shaping the answers and which stopped. Your page under "Dropped" often means not a drop in quality but that a competitor's fresh material displaced it — and then the work is with that material, not with your page.

Don't try to close every gap at once. Sort them by size and by query type — a commercial "where to buy / order" beats an informational one — and start with the top 3. Every closed gap adds to your Share of Voice instead of a vague "we're losing".

Review the whole ranking every few weeks — the AI engines keep finding new players on your queries. A new name near the top usually means a competitor just entered your niche or started publishing heavily: an early signal worth checking in their card.

Sentiment is routinely underrated. If your Share of Voice is 20% but a sizeable slice of the mentions is negative, the AI engines are talking about you in a bad light, and the customer will pick the competitor with the smaller share but the better image. Handling negativity is as much a part of GEO as growing mention counts.

Use filters as a diagnostic tool. An aggregate SoV may look average, but isolate a single cluster or provider and a drop or spike becomes obvious. It produces far sharper decisions than working off pool-wide averages.

Don't try to be everywhere. Take the top of the list, filter to media and communities — those are the venues you can actually influence and that AI already cites on your queries. Competitor domains from the same list are useful differently: their pages show the format of material the engines are willing to cite in your niche.

Check "Winners & Losers" weekly — it's the fastest way to spot a shift before it shows up in the headline metrics. Someone else's fresh listicle appearing under "Gained" usually means a new ranking shipped in your niche and it's worth getting into; your own URL under "Dropped" is a reason to look at what changed on the page or at competitors.

Compare "Your brand cited" against the competitors' counts in "Cited from this source". A high citation share for the outlet with a zero for your brand is the strongest argument to go there with a publication: the engines already trust the source, you're simply not in it.

Before negotiating a placement, open the target page at this level: if your brand is already in it but competitors are the ones being cited from it, the problem isn't the outlet — it's how you're presented inside the material. Sometimes updating the mention beats shipping a new publication.

Start from the top rows by "Found": a page the engines saw ten times and never cited pays off more than a perfect new publication. It already has the appearances — all it lacks is wording the model can quote.

Don't rewrite the page "in general". The edits target specific uncited queries — paste in the ready paragraph that answers the prompt's intent directly, and the page starts answering the exact query AI was skipping it on. Closing query after query is how domain citability grows systematically.

Clear the technical barriers first, then work on content. If an AI bot is blocked in robots.txt, the engine simply can't read the page — and even perfect copy won't reach the answers. Bot access and llms.txt are the foundation that page improvements and publications rest on.

Don't treat external publications as an add-on to on-site work. For many niches this channel lifts citations faster than anything you change on your own site: AI already trusts these sources, you just need to be inside the right material. Editorial placements are the strongest — one article on an outlet the model already cites quickly raises the brand's share of mentions.

Work strictly top to bottom and don't try to close the whole list. The first few steps deliver most of the gain — spending effort on the tail before the top is done isn't worth it. And don't rank steps by how prestigious an outlet feels: the plan is already sorted by how much new visibility each publication brings.

An article from an idea is a starting point, not a publish-ready piece. Fact-check the claims and figures and adapt the text to the outlet's requirements before sending it to editors. And the fuller the brand knowledge base, the less the model has to improvise — fill it in before the first generation.

Load the key facts up front — numbers, pricing, advantages, differences from competitors — and re-index a source whenever the page's data changes. Then every generated article speaks about the brand with the same accuracy, and you won't have to fix facts by hand after each generation.

This is the metric that makes GEO measurable: recommendation → publication → URL in a campaign → citations in the next monitoring cycle. Mind the lag: it takes 2–3 weeks on average (depending on the outlet) before a piece starts being cited — the engines need time to index it and pull it into their sources. If a URL still gains no citations after that, rework the material or try a different outlet.

Start working on negativity from the "Sources" block, not the answer itself. Open what the model referenced: it's often an old review or a competitor's roundup. Closing the negativity means giving the engine a fresh, more authoritative piece it will cite instead of the outdated one.

A competitor advertising on your branded or comparison prompt is a signal the topic is already being monetized. That's a cue to strengthen your organic presence there: the more confidently the AI answers with your brand, the less weight the paid block beside it carries.

Focus not on one-off impressions but on advertisers with a high prompt count: if a single domain buys ads across many of your topics, that's your systemic competitor for paid traffic from AI engines.

Use the section as reconnaissance on the paid channel: domains that consistently buy "Promo" on your queries reveal which topics AI engines already treat as commercially valuable — and where growing your organic visibility pays off most.

The report is the dashboard translated into the language of decisions: instead of "SoV 15.6%" it tells you the brand leads in acquiring but slips in media and communities, and what to do next. Use it as the agenda for your team's weekly AI-visibility sync.

An AI strategist isn't a lightweight alternative to the dashboard. It's a layer of deep work on top of the data: the work a senior marketer or external consultant typically spends days on gets done in one evening of conversation. The dashboard shows cuts, the command center sets priorities, the AI strategist does the analytics and strategy.

For PAT access, add `--header "Authorization: Bearer gs_<your-token>"` to args. On Windows, there's a known mcp-remote quirk with space escaping — set the token via an `AUTH_HEADER` env var (value `Bearer gs_...`) and reference it in args as `Authorization:$AUTH_HEADER`. Exact snippets are on the /mcp page.

To connect the server for one project only, drop a `.cursor/mcp.json` file with the same config in the repo root — Cursor picks it up automatically and the connection won't leak into other projects.

WebMCP complements OAuth/PAT rather than replacing them. For writes, scripts, and external clients (Claude Desktop, Cursor) use a token connection; WebMCP is the quick read-only strategist in the same window where you already work.

Don't guess tool names — describe the task. The agent picks from the available MCP methods and chains them as needed. If a follow-up confuses the agent, rephrase: "Based on monthly monitoring data, …"

The strength of MCP is exactly this — asking "why" and "what if". The dashboard shows how things stand; MCP lets you dig for causes and find non-obvious connections. One conversation with the agent typically replaces 1–2 hours of manual dashboard work.

Brief quality depends on how you ask. Good templates: "give me 5 key entities", "suggest schema blocks", "find 3 facts about my brand from monitoring that would strengthen citation". The more specific the request, the denser the result.

The full list of available tools with descriptions is on the /mcp page in the dashboard. Need a tool or data cut that isn't there yet? Message support — the MCP toolset is actively expanding.

To prevent this from happening again, add all possible brand spellings in advance at Settings → Brand → Name variations.

Merging is irreversible — carefully verify the selected competitors really are the same brand before confirming.

Your feedback improves the processing algorithms. Every flagged response is reviewed by the team and recalculated when needed.

On lower plans, focus on niche prompts — queries that describe your service or product as precisely as possible. On higher plans, add broad prompts for a market overview — that's how you see the full competitive landscape in AI.

Add every spelling of your brand at Settings → Brand. This is critical: if AI writes the brand differently — e.g. "T-Bank" instead of "Tinkoff", or "SberBusiness" instead of "Sber" for a separate product line — the mention will be missed without the variation.

Competitive landscape quality depends directly on prompt quality. Generic prompts give a broad market overview; niche ones surface precise competitors. Combine both types and group them into clusters — even rough grouping gives the system context for more accurate recommendations.

A high share of positive mentions and recommendations is a sign of strong expert reputation. If sentiment is negative, check what's said about you in the sources AI uses (Wikipedia, reviews, forums).

Start with the most frequent and commercially important gaps. Creating content on those topics is the fastest way to increase Share of Voice.

Read Share of Voice and Citation Share together. SoV high but Citation Share low — you're mentioned but AI links to competitor sites: invest in domain authority (pillar pages, schema, external mentions of your own site). Citation Share growing faster than SoV — the brand isn't on everyone's lips yet, but your site is already winning the battle for AI trust.

The priority of the same recommendation can shift from cycle to cycle: a competitor publishes a new piece — priority rises; the gap narrows — priority drops. Don't try to close everything at once: today's top-of-feed action may slip tomorrow, and an irrelevant signal may fall off on its own.

If you work with an agency, give them access to GEO Scout — so both sides work with the same objective data.

Your referral code is generated from your account name plus a unique suffix. The code can't be changed, but it's easy to remember.

Commission applies to every payment, including renewals and plan upgrades.

Verify your payment details before requesting a payout — they're captured at the moment the request is created.