Tools to Monitor Brand Mentions in AI Answers from ChatGPT, Gemini, and Perplexity (Multilingual)

Tools to monitor brand mentions in AI answers from ChatGPT, Gemini, and Perplexity, multilingual, have to let you set where the prompt runs and write the prompt in the language the buyer uses. That is the whole trick. We do not brief a magic language pack. We have not documented a Promptwatch SKU that "supports 50 content languages," and we will not invent one for a slide. What we do use is country, state, and city tracking on prompts we typed, plus topics, tags, and personas. The language is the prompt text. The location is the targeting lever. Multilingual work is both of those on the same row, not a checkbox on a pricing page.

The reason this needs saying out loud is that the market keeps asking for a language pack and the vendors keep implying one exists. A language pack would be a feature that says "we monitor in 50 languages," and it would be a checkbox a buyer could tick and stop thinking. The honest version is that there is no language pack. There is a prompt you type in a language, and a location you set for that prompt. The work is in the typing and the setting, not in a tier you buy. A team that waits for a language pack to ship before doing multilingual work waits forever, because the feature they are waiting for is two things they already have.

The monitor we run is Promptwatch. Paid plans include ChatGPT, Gemini, and Perplexity daily from the UI. Essential is $95 a month with country-level targeting. Professional is $245 a month and adds state and city targeting. Explore is ChatGPT only and limited to 10 prompts, which makes it a sandbox rather than a program. Kick-off is $199 a month with state and city targeting, unlimited projects, and 10 seats, which is the plan we reach for when more than one market brand sits in the account.

The plan ladder maps to how serious the multilingual program is. Explore is the sandbox, ten prompts on one engine, which is enough to learn the UI and not enough to run a market. Essential is the first plan that runs a market, with country targeting and the three engines, and it is where a single-market brand lands. Professional is the plan that runs a market with local depth, because state and city targeting is what lets you see whether a local pack or a city-named vendor list is showing up in the answers. Kick-off is the plan that runs multiple markets at once, because unlimited projects is what lets you keep the markets separate instead of blending them into one account-wide average.

How we run multilingual without inventing a language pack

List the markets that close deals. Not the countries on a homepage map. The ones with pipeline. A market with no deals behind it does not earn a row, because every row costs a prompt and a weekly review.

The markets-with-pipeline rule is the one that keeps the program honest. A homepage map shows every country the brand ships to, which is a marketing artifact and not a sales one. A pipeline list shows the countries where deals closed, which is a sales artifact and not a marketing one. The rows in the monitor should match the pipeline, because a row that matches a market with no deals is a row that gets reviewed every week and never produces a decision. The cost of that row is not just the prompt. It is the half hour of review time spent on a number that does not change a decision.

Write the prompt in German if the buyer asks in German. Duplicate the English prompt as a second row. Tag them so the report does not average the two into a meaningless number. Set country, and set city when a local pack or a city-named vendor list shows up in the answers. Do not assume a US ChatGPT answer matches a DACH Gemini answer, because the two engines read different corpora and serve different defaults.

The duplicate-and-tag step is the one that prevents the most common reporting error, which is averaging across languages. An English prompt and a German prompt for the same buyer question produce two different mention rates, because the two engines read different corpora and the two languages have different competitive sets. A report that averages the two into one number reports a figure that is true in neither market. The tag is what keeps them separate on the report, and the tag is the part that makes the number mean something when it lands in a review.

Mention versus citation still splits across languages. Perplexity footnotes more than ChatGPT, so a Perplexity row will look more cited even when the brand is equally present in both. A Reddit citation will not match your market just because you set Germany on the row, because the thread that fed the answer might be in English or in a different regional dialect. The targeting sets where the assistant answers from. It does not set the language of the sources the assistant pulls.

The mention-versus-citation split is the part that catches a team that compares engines apples to apples. Perplexity footnotes more than ChatGPT by design, so a Perplexity row will always look more cited than a ChatGPT row even when the brand is equally present. A team that reads the Perplexity number as a win and the ChatGPT number as a loss is reading an engine behavior as a market result. The Reddit citation mismatch is the other half of the same trap. A Germany-targeted prompt can still pull a Reddit thread in English, because the assistant pulls from the corpus it has, and the corpus is not bounded by the targeting. The targeting sets where the assistant answers from, not what it reads.

GSC generative reports have a country dimension for Overview impressions. That is Google. It is not ChatGPT in another language. Keep both, and keep them labeled differently, because a German Overview impression and a German ChatGPT mention are not the same unit. See the AI features reading.

The GSC country dimension is the part that gets confused with the Promptwatch country targeting, because both say "country" and both produce a number by country. The GSC number is a Google Overview impression, which is a Google surface measured by Google. The Promptwatch number is a ChatGPT or Gemini or Perplexity mention, which is a non-Google surface measured by Promptwatch. The two numbers are not the same unit, and a slide that drops them into the same column produces a column where Germany looks twice as visible as it is, because the two numbers got added. The label is what stops the addition.

Other tools in a multilingual brief

Otterly at $29 covers the cheap on-ramp. Gemini and Claude are add-ons. The data can lag up to seven days, and we inherit that lag. We will not run a DACH launch review on a lagged Gemini add-on, because the review would report last week's answer to a stakeholder who already moved on.

Otterly is the cheap on-ramp and it is the one we are clearest about not using for a launch. A launch review has to report this week's answer, because a launch is a this-week event. A seven-day lag turns the launch review into a history report, and a history report is not what a stakeholder wants the week after a launch ships. The lag is fine for a steady-state program where the week's delay does not change a decision. It is the wrong tool for a launch, and the price is not worth the wrong timing.

Peec at $95 covers three models, with extras running $35 to $165 and Claude gated to Enterprise. It is fine if the client already lives there. It is still not a language pack, and the add-on math is the part that bites when a market list grows.

Peec is the same price as Promptwatch Essential and it is the one we are clearest about not migrating a client off of. A client who already lives in Peec has a workflow built around it, and a migration costs more than the workflow is worth. The add-on math is the part that bites when the market list grows, because each extra model is an extra line item, and a five-market program on Peec can stack add-ons past the Promptwatch Professional price without ever getting the crawler logs. The math is the reason we do not start a new multilingual program on Peec, even though we do not force a client off it.

Profound Starter is $99 annual, ChatGPT only. Growth is $399 annual. Enterprise runs about $2,000 to $5,000 or more per month. Read the engines before you add "multilingual" to the Profound slide, because the Starter tier will not monitor Gemini or Perplexity in any language.

Profound is the one where the tier matters most for a multilingual brief, because the Starter tier is ChatGPT only and ChatGPT only is not multilingual in any useful sense. A multilingual program needs at least the three engines, and the three engines on Profound are not on the Starter tier. A team that puts Profound Starter on a multilingual slide is putting a single-engine tool on a multi-engine brief, and the slide will not survive the question "how does this cover Gemini."

Ahrefs Brand Radar, at $199 a month plus the plan, and Semrush AI Toolkit, at $99 a month per domain with modeled prompts, are on the bench. We keep them. Modeled English-heavy corpora are a poor substitute for prompts you wrote in the buyer's language, because the model is guessing what a German buyer would ask rather than reading the question a German buyer typed.

The bench tools are the ones we keep but do not lead with. Ahrefs Brand Radar and Semrush AI Toolkit are modeled, which means they generate prompts from an English-heavy corpus and project them onto other markets. The projection is a guess, and a guess is not a measurement. A German buyer types a German question, and the modeled prompt is the model's idea of what a German buyer would type, which is a different string. A program that runs on modeled prompts runs on the model's imagination of the market, not the market itself.

Professional is $245 a month with 25 million crawler logs. Business is $579 a month with 100 million. Agent Analytics is also included on the agency plans, while visitor analytics through a script or GTM starts on Essential. How to get cited by ChatGPT still applies once the page is fetchable in that market, because a page the bot cannot fetch cannot be cited in any language.

The crawler log tier is the part that turns a mention program into a diagnosis program. A mention program tells you whether you are visible. A crawler log program tells you whether the bot fetched you, which is the question that separates a content miss from a technical miss. A multilingual program without crawler logs is a program that diagnoses every miss as a content miss, because it has no fetch record to split the cases. The fetch record is what makes the diagnosis honest, and the fetch record starts on Professional.

What we will not do

We will not invent a 50-language feature. We will not blend US and DACH into one mention score. We will not auto-translate prompts and call that localization, because a machine translation of an English prompt is not the question a German buyer would type.

The three things we will not do are the three shortcuts that look like multilingual work and are not. A 50-language feature is a slide line, not a program. A blended US and DACH score is a number that is true in neither market. An auto-translated prompt is a string a German buyer would not type, run against an engine that reads German, which means the engine answers a question nobody asked. Each shortcut produces a deliverable that looks like multilingual work and reports a number that does not reflect the market. The discipline is to do the typing and the setting, which is slower and produces a number that is real.

FAQ

Does Promptwatch auto-translate prompts?

Not as a feature we brief. You write the language.

Is country tracking the same as multilingual?

Country is the location lever. Language is the prompt text. You need both when the market is not US English.

Can Explore cover Gemini in Spanish?

No. Explore is ChatGPT only. Paid Promptwatch is the three-engine start.

What to do this week

1. List markets that closed a deal in the last two quarters.
2. Write 10 prompts per language you sell in.
3. Set country on each row in Promptwatch Essential ($95). Use Professional or an agency plan if state or city targeting matters.
4. Keep GSC countries for Overviews only.
5. Email hello@1001seomedia.com with the market list if you want us to build the rows.