How can I tell whether AI names competitors instead of my brand?

Direct answer

The measurement runs a bank of category questions that name no brand, repeated across several models, recording per answer which brands appear. In the 4-week window closed on September 14, 2026 I measured 21 markets of e-commerce platforms that way and three different leaders came out: Shopify first in 18 markets, Magento in Italy and Germany, Tiendanube in Argentina. The count per market and per language is what shows when a different brand holds the place.

Step 1. Write the category questions without naming the brand

What to look at: the questions a customer asks while still choosing, written with no brand inside them. A question that already names the brand returns an answer about that brand and leaves the comparison out, which is the part worth measuring here.

Which module: Prompt Simulation generates customer prompts and detects with fixed rules whether the product appears in the answer.

Step 2. Repeat the same question across engines and across runs

What to look at: the same question returns different lists across models, and different lists across repetitions of one model. A single run describes one draw.

Which module: Prompt Tracking runs prompts across the five engines: ChatGPT, Gemini, Google AI Mode, Perplexity and Claude.

The variation runs over time as well. According to a study by Semrush, a company that sells SEO software, across more than 230,000 prompts between July and October 2025, domain citation rates moved sharply week to week¹. A one-day measurement stays tied to that day.

Step 3. Count mentions and answers separately

What to look at: two numbers that look alike and say different things. An answer can name a brand more than once: mentions count every time it appears, and answers count each answer once. A brand's mention share is its part of the category's verified mentions in the period.

A brand named three times inside one answer and a brand named once across three answers add up the same in mentions and split differently in answers.

Step 4. See which brand holds the place

What to look at: the list of brands an answer names when the brand itself stays out. That set is the competitor as the AI measures it, and it can differ from the competitor in the commercial plan.

Which module: AI Visibility Gaps identifies where AI cites the competition and leaves the brand out. It combines real structured data with per-page signals.

Step 5. Separate a clear lead from a tie

What to look at: the margin of error on the share. I calculate the 95% interval with the Wilson method, over mentions and over answers. I publish a lead as clear when the second brand's share falls outside the leader's interval in both calculations. If it falls inside, it is a statistical tie.

A tie read as a lead turns noise into a plan.

Step 6. Cut by market and by language

What to look at: the same category measured in two markets returns two rankings, and measured in two languages it does the same. A global average covers both.

Which module: AI Market Leaders runs once per ISO week per domain.

The field's critical survey sets the expectation for this cut: topical relevance and position within the context are the most reproducible findings, and generic heuristics transfer poorly across contexts². A finding from one market holds for that market.

Step 7. Look at the sources behind the answer

What to look at: the domains an answer cited when it named the competition. OpenAI documents that its web search tool returns a field with every URL retrieved, wider than the citations that end up displayed³, so the visible list is a subset of what entered the context.

On the site's own side of the measurement, Bing Webmaster Tools reports to a site owner the total citations, the daily average of cited pages, the grounding queries the AI used to retrieve the content, and citation activity per URL⁴. And a 2026 work is a reminder that a citation and a contribution to the answer are two separate measures⁵.

Which module: Prompt Tracking shows, for each prompt and model, the sources the answer cited, and AI Market Leaders groups them by brand.

What my data shows

Three published measurements serve as examples of what these steps detect.

The first is a cross by market. In the 4-week window closed on September 14, 2026 I compared 21 markets of e-commerce platforms across 1,150 answers from ChatGPT, Gemini and Claude. Shopify came first in 18 of the 21 markets, with 18.5% of the mentions in the US. Magento took the top spot in Italy and Germany. Tiendanube led in Argentina, with 16.9% of the mentions in that market. The full measurement is at Shopify leads e-commerce platforms in 18 of 21 markets.

The second is a cut by language. In the same window, in the global market and with 108 answers per language, Mercado Pago held 3.0% of the mentions in payments and fintech platforms when the three models answered in Spanish, with 21 mentions and presence in 16.7% of the answers, and it had no mentions when they answered in English. The full measurement is at Mercado Pago in payments and fintech platforms.

The third is one category inside a single market. In real estate portals and software in Mexico, across 111 answers from ChatGPT, Gemini and Claude in Spanish in the same window, Inmuebles24 held 13.8% of the mentions and Vivanuncios 7.4%, with the leader named in 71.2% of the answers. The distance between first and second is a clear lead: Vivanuncios' share falls outside Inmuebles24's interval, which is the test from step 5. The full measurement is at Inmuebles24 leads real estate portals and software in Mexico with 13.8% of mentions.

All three are the same finding seen from three cuts: the brand that holds first place changes with the market, changes with the language of the question, and inside one market the distance to the second says whether there is anything to chase. A measurement without those cuts averages over markets where first place belongs to a different brand.

How to tell whether AI is leaving you out

1. Write down which brands hold your place when you are missing. The list often surprises you, because the competitor the AI assembles rarely matches the one in the commercial plan.

2. Before reacting to a gap, check whether it is real. If the second brand sits inside the first one's margin of error, that is a tie, and chasing it costs you work without giving anything back.

3. Repeat the measurement per market and per language before drawing conclusions. The three examples above show the leader changing with both, so a single global figure tells you little about the market you care about.

4. When you find a question where you are missing, start by reading what the brand that does appear answers. It is worth doing that before writing, because the difference is almost always that their page answers the question while yours talks about something else.

Frequently asked questions

¿Cómo detecto las menciones de mis competidores frente a las de mi marca en las respuestas de IA?

Both are counted in the same question bank and the same window: for every answer the brands that appear are recorded, and from there come the brand's own share and each competitor's. Comparing two measurements taken in different windows mixes the period's effect with the brand's.

¿Cómo sé si la IA nombra a un competidor donde debería nombrar a mi marca?

The indicator is the portion of the category's answers where a competitor appears and the brand itself stays out. It reads alongside mention share: a brand can hold few mentions and be present in many answers, or the other way around.

¿Cómo identifico qué competidores menciona la IA?

The list comes from the answers themselves, with no prior list as a starting point: every brand named is recorded and then reviewed one by one for whether it is a real competitor. A name that appears inside another word inflates the count, so the match has to be on the whole word.

¿Qué hace falta para detectar menciones de competidores en las respuestas de IA?

A fixed question bank, repetition per model and a reviewed brand registry. The rest is counting.

How I measured it

AuraMetrics public benchmark: it measures every week which brands ChatGPT, Gemini and Claude name across 20 markets and 15 industries, in Spanish and in English. The models are queried through the API with web search off. Every mention is counted against a curated brand registry, with whole-word matching. Every brand that appears in a Quick Insight has human review.

The three measurements cited come from the 4-week window closed on September 14, 2026: 1,150 answers in the cross of 21 e-commerce platform markets, 108 answers per language in payments and fintech platforms in the global market, and 111 answers in Spanish in real estate portals and software in Mexico.

Declared scope: the benchmark queries the models through the API with web search off, so its results describe what a model answers from its own knowledge and its internal retrieval, and they leave out what a browsing search product returns. The category and market cut from the Observatory of brands in AI joins this guide with the v2.2 revision.