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.

Your first report is built from a single run — it's a starting point. The longer monitoring runs, the richer the reports get: period-over-period dynamics, competitor comparisons, and the measurable impact of your changes.

Start with a few lower-cost engines and add premium ones (Claude, Yandex with Alice) selectively — where that specific model matters. This keeps coverage high while keeping spend under control.

After a rebrand, keep the old name in the variations list for at least a year: LLMs train on historical data and keep mentioning the brand by its previous name long after. Remove the variation — lose those mentions and a real chunk of your Share of Voice.

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 3–5 days to accumulate data. More observations means sharper signals about real patterns and weak spots, not random noise from one or two AI answers.

Start your analysis with dynamics, not absolute values. A 13% SoV is excellent if it was 5% a month ago and alarming if it was 25%. Always look at the change indicator vs the previous period.

Don't try to close every gap at once. Sort by prompt frequency and intent priority (commercial "buy / order" beats informational), and start with the top three. Each closed gap is +N mentions to your Share of Voice.

Sentiment is often underrated. With 20% SoV but 40% negative mentions, AI is mentioning you in a negative tone and the customer picks a competitor with lower SoV but a positive image. Working with negative sentiment is as much a part of GEO as growing the mention count.

Review the ranking in full every few weeks — AI keeps surfacing new competitors on your queries. Enable the ones truly in your market and strip out the noise. A cleaner list means sharper analytics and less noise in your KPIs.

Closing a gap is a concrete target list, not "make the site better". Open the gap, look at whose domains are cited (usually competitors plus catalogs and media) and the Fan-Out sub-queries — and you get a precise plan: which venues to appear on and which phrasings to write for.

Don't try to "be everywhere". Start with the media and communities at the top of the list — venues you can actually influence and that AI already cites for your queries. The expanded competitor URLs tell you the exact content formats that make it into answers.

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 rewrite the page "in general". Answer capsules target a specific uncited query — move the ready paragraph and markup over, and the page starts answering exactly the intent AI was skipping it for. Closing query after query, you build up the domain's citation rate systematically.

Close the technical barriers first, then optimize content. If an AI bot is blocked in robots.txt or a page has no markup, even perfect text won't make it into answers. Schema, llms.txt, and open bot access are the foundation that answer capsules and publications run on.

Don't treat external publications as an add-on to on-site work. For many niches this channel grows citations faster than on-site changes: AI already trusts these sources — you just need to land in the right material. Editorial publications are especially strong — one article on an outlet the model already cites quickly lifts the brand's share of mentions.

Don't order "an article about the topic in general". The content plan is built from what AI already cites, so an article from it is more likely to land in answers. Keep 1–2 angles per query (e.g. comparison + guide) — that covers more phrasings without spreading thin.

Load the key facts into the knowledge base up front — numbers, pricing, advantages, differences from competitors. 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 usually takes 2–3 weeks (depending on the outlet) before the material starts getting cited by the AI engines — they need time to index it and pull it into their sources. If it still picks up no citations by then, rework the material or try another 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.

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" vs "Tinkoff"), 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 high_impact may be tomorrow's quick_win, 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.