Common setup mistakes

Mentions are matched against your domain’s brand name and domain name. A legal name (“Acme Software Ltd.”), an old name after a rebrand, or a product name that differs from the company name means real mentions are not counted as yours.

Check. The Brand column on the Domains page, against how answers in prompt detail actually name you.

Fix. Use the short name people and engines use. To change it after setup, ask a connected agent to update the domain, or contact us. See add a domain.

Every prompt, competitor and tag belongs to one domain. If your product has its own site, tracking it under the corporate domain mixes two brands’ results.

Check. Whether the prompts on one domain are really about two different products.

Fix. Add the second site as its own domain (Pro allows 5, Premium 50) and move its prompts there.

With 5 prompts, each prompt is worth 20 points of visibility. One answer changing between runs swings the score by 20 points, which looks like a trend and is not.

Check. The tracked prompts count on the Dashboard.

Fix. Track enough prompts that one flip is a small move: 40 prompts puts each at 2.5 points. See choosing prompts.

Prompts that contain your brand name (“is Acme good for invoicing?”) will almost always mention you. A set made mostly of them reports high visibility that says nothing about whether buyers who do not know you yet will find you.

Check. Count how many prompts include your brand or product names.

Fix. Keep branded prompts in their own tag and make most of the set unbranded, buyer-intent questions.

A broad prompt next to your category (“best CRM” for a help desk tool) will never mention you. It lowers visibility forever and is not a finding.

Check. Read the answers to your lowest-performing prompts. Are the brands named even your competitors? Fix. Delete those prompts and replace them with the questions your buyers actually ask.

New prompts start as not ranked, so a big batch lowers the score even when nothing else changed. See why did my score drop?.

Fix. Tag each new batch, and compare periods using the tags that existed in both.

One AI response costs 0.015 credits. A prompt run daily on 8 engines is 8 responses a day, about 3.6 credits per 30 days. The same prompt run hourly is 192 responses a day, about 86 credits per 30 days: most of a Pro plan’s 120 monthly credits for one prompt.

Check. Each prompt’s schedule on Prompts.

Fix. Use daily for most prompts and weekly for long-tail ones. Save hourly for a few prompts you are actively changing. See run frequency and credits.

Every prompt runs from one country (default US), and the country sets the query language. A German shop tracked from the US gets English answers that rarely name local brands.

Check. The country on each prompt.

Fix. Track the same prompt text in each market you sell in. See countries and languages.

The onboarding step “AI providers” starts with every engine selected. Unticking one to try things out means those prompts never run on it.

Check. The engines on a few prompts, from Prompts.

Fix. Clear the engine list so prompts use every enabled engine, including new ones added later. See no data for an engine.

Competitors are not part of onboarding, so new workspaces often have none. Share of voice still shows every brand named, but the competitor views and comparisons stay empty.

Fix. Add your main competitors on Competitors, with their domains. See competitors.

Without tags you cannot split the numbers by product line, funnel stage or branded vs. unbranded, and every comparison mixes them.

Fix. Give each prompt one tag. Every dashboard figure can then be filtered by it.

The citation gap report, query coverage and every organic report need Search Console or Bing Webmaster Tools data. Without it they are empty, which is easy to misread as “no gaps”.

Fix. Connect them under Data Sources. See Search Console and Bing.