How Mencoro works

Mencoro shows how AI engines and Google present your brand next to your competitors, for the questions and keywords you choose, in the countries you choose. This page explains what we collect and exactly how every number in the app is calculated.

We track three kinds of results:

• AI answers from ChatGPT, Perplexity, Google AI Overview and Google AI Mode: what they say and the sources they cite.
• Google SERP: the top 10 organic results.
• Google Shopping: the organic product listings.

You read the results in the app, through the REST API, or in your AI assistant through the MCP server. The live demo shows it all with sample data.

From query to data

What you set up

A project is one brand. In it you define:

• Your brand: its names (“Coca-Cola”, “Coke”) and its domains.
• Competitors: each with its own names and domains, as many as you need.
• Queries: AI prompts for the AI engines and SERP keywords for Google SERP and Shopping. You can write them or generate suggestions.
• Engines and countries: each query runs separately on every engine and in every country you pick. Each combination is a variant.
• Check frequency: daily, weekly or monthly, per query.
• Passes per check: 1 to 3, on AI engines only. Each pass asks the same question again, to smooth out answers that vary.
• Groups: sets of queries to read results by topic.

How a check runs

A check is one run of one query on one engine in one country.

1. Every hour, Mencoro picks the queries that are due.
2. It asks the engine through its public interface, located in the query’s country. The engine receives only the query text and the country.
3. In AI answers, a language model finds every brand named and classifies it, and each mention is checked against the visible text. SERP results are matched to brands by domain, and Shopping listings by seller.
4. The results become daily figures, visible in the app shortly after.

What each type of result includes

• AI answers: the text and its sources. ChatGPT is asked with web search on, and its sources include the pages its search retrieved even when the text does not cite them. Google shows an AI Overview only for some searches; when there is none, the check has no answer. Google AI Mode is not available in France or Cuba.
• Google SERP: the top 10 organic results. Ads and other modules do not count.
• Google Shopping: the organic product listings. Sponsored listings do not count.

Location is set by country and the language is the country’s main language. There is no city or device targeting.

Mentions and links

What counts as a mention

A brand mention is a brand named in the text of an AI answer: a company, product, service, app or website.

• It counts in plain text, and as the text of a link that presents the brand.
• It does not count inside a URL or a domain (that is a link), as a source attribution (“according to…”), or as an indirect reference (“the company”).
• It goes by identity, not exact spelling: “TimesRoom Barcelona” counts for “Times Room BCN”. A translated name does not.
• Every occurrence counts: a brand named three times has three mentions.

How each mention is classified

Type describes structure and tone describes judgment: advice to avoid a brand is a recommendation with a negative tone.

Links

A link is a source the answer returns. It is yours when it points to one of your domains or their subdomains, even if the answer never names your brand. Its position is its place in the list of sources. Links feed Link position and Share of voice, but not Coverage or Favorability, which are about what the answer says.

Metrics and formulas

Coverage

Coverage = checks in which you were named ÷ checks that returned an answer × 100

• Each query on each engine and country counts once per day it was checked.
• With several passes, named in 2 of 3 passes counts as ⅔.
• Checks with no answer, such as a search with no AI Overview, are left out.
• A link to your site alone does not count.

Example. 10 prompts on ChatGPT, checked daily for 7 days, are 70 checks. Named in 21 of them, your Coverage is 30%.

Favorability

Favorability = (positive + ½ × neutral) ÷ all mentions × 100

Example. 12 positive, 6 neutral and 2 negative mentions give (12 + 3) ÷ 20 × 100 = 75. A score of 50 means your mentions balance out to neutral, which is common in factual lists. It is empty when there are no mentions, and it is calculated the same way per engine and for each competitor. In the API and the MCP server, it is the positivityIndex field.

Share of voice

Share of voice weighs mentions instead of counting them, because a recommendation is worth more than a passing reference in a long list.

weight = type × tone × qualifier

Share of voice = your weight ÷ (your weight + tracked competitors’ weight) × 100

• A link to your domain counts as a direct, neutral reference: 0.14.
• Only you and your tracked competitors count. Without a tracked competitor present, it stays empty rather than showing 100%.
• It is calculated per day, engine and country. Over a date range it is the average of the days in which you appear, so read it together with Coverage.

Example. You are recommended positively (1.0) and your site is cited (0.14). Competitor A is listed neutrally (0.4 × 0.7 = 0.28) and competitor B is compared negatively “for large teams only” (0.5 × 0.1 × 0.7 = 0.035). Your share of voice is 1.14 ÷ 1.455 = 78.4%, where counting brands would have given 33%.

Positions and stability

• Mention position: where you are first named in each answer, averaged across passes and days, then across queries. Days you are not named are left out: position says where you rank when you appear, Coverage says how often.
• Link position: the same, for your first link among the sources.
• Rank distribution: your queries grouped by average position (1, 2–3, 4–9, 10–19, 20–29, 30+).
• Stability: how much your daily position varies, as a standard deviation. It needs two days of data. The query table uses the last 14 days.

Google SERP and Shopping

• SERP position: your best result in the top 10, matched by domain. A result on another site that names you does not count. SERP coverage: the share of checks in which you rank.
• Shopping position: your best listing, where a listing is yours when you are the seller. Shopping coverage: the share of checks with one of your listings.
• Tone, type and share of voice only apply to AI answers.

How figures add up

• Days are UTC days. History is kept for 16 months.
• Passes are averaged, so more passes give a steadier figure, not a bigger one.
• Trend arrows compare with the previous period of the same length. For positions, lower is better.
• Weekly charts start on Monday, and monthly charts use calendar months.
• A query in several groups counts in each of them.
• The project page and the aggregated query view cover your date range. The detailed query view shows each query’s most recent position.

Updates and usage

When data updates

• Mencoro checks for due queries every hour. A query is due once its interval has passed since the hour its previous check ran: 24 hours, 7 days or 30 days. So a daily query runs at about the same hour every day, and “monthly” means every 30 days, not on a set date.
• A new query runs its first check straight away, if your plan has checks left.
• You can force a check at any time. Forcing every query in a project is limited to once every 30 minutes.
• Paused queries do not run.

Checks and your plan

Plans differ only in checks per month, and every plan includes unlimited projects, competitors and users. Each pass uses one check; SERP and Shopping always run one pass.

checks per month = queries × engines × countries × passes × runs per month

Runs per month are 30 for daily, 4 for weekly and 1 for monthly. Example: 25 prompts on 3 AI engines in one country, 1 pass, weekly: 25 × 3 × 1 × 1 × 4 = 300 checks.

• Checks are used when they start, so one with no answer, or one that fails, still uses its checks.
• When your checks run out, the queries that have waited longest go first and the rest wait for the monthly reset.

See pricing for the plans.

Daily, weekly or monthly?

Weekly suits most queries: AI answers vary from run to run, and a weekly sample shows real changes without spending checks on noise. Use daily for launches, PR issues or live campaigns, and monthly for slow categories or to cover more queries with the same checks. Covering more queries usually teaches you more than checking fewer more often.

Reliability and limits

Detection accuracy

We measure how well brands are detected against a set of real AI answers reviewed and corrected by a person. Accuracy is the share of brands detected and assigned to the right owner, shown in the app for your tracked brands and for untracked competitors. Tracked brands score higher because the system knows exactly what to look for. If you spot a mistake, report it from the answer.

What the data is and is not

• The queries are yours, not your customers’. Real conversations with AI assistants are private, so Mencoro shows how engines answer the questions you choose, not what every person sees.
• Answers vary. The same question can get another answer on another day. Passes and repeated checks reduce the noise; trends tell you more than a single answer.
• No personalization. A check carries no account, history or personal context.
• Privacy. Engines receive only the query text and the country. Nothing from your account is sent to them.

The figures are directional: they show whether you are gaining or losing ground, where and against whom, not an absolute count of how many people see your brand.

FAQ

Why does a Google AI Overview check come back empty?

Google only shows an AI Overview for some searches. When it shows none, the check returns no answer: it still uses its check, but it is left out of Coverage, because there was no answer in which your brand could appear.

Why is my share of voice empty?

Share of voice compares you with your tracked competitors, so it only appears once at least one tracked competitor is mentioned or linked in the same answers.

My site was linked but my brand was not named. Does that count?

As a link, yes: it counts toward Link position and Share of voice. As a mention, no: Coverage and Favorability only use your brand named in the text of the answer.

Why does a query show a different position in the query table and on the project page?

The project page and the aggregated query view average the date range you select. The detailed query view shows the most recent position recorded for that query. One is a trend, the other a snapshot.

Does adding or removing a competitor change my past data?

Adding one only affects checks from then on. Removing one deletes its mentions, so your past share of voice is recalculated without it.

Why did ChatGPT give an answer with no links?

Mencoro asks ChatGPT with web search enabled, but the model decides whether to search and cite. When it answers from what it already knows, there are no sources to record. Mentions in the text are still counted.