Rule: Mergado Pilot#

Mergado Pilot is a set of rules that improve the content of your feed in bulk using AI and segment products by their performance in advertising or in your online store. It currently offers six types:

• Title and description improvement (Element enhancer in the interface) — rewrites or enriches the product title (Title) or product description (Description).
• Q&A (question and answer) — generates question/answer pairs for Google Shopping.
• Product highlight — generates key product features for Google Shopping.
• Custom prompt: modifies the selected element exactly according to your instruction, without a preset system prompt. Intended for advanced users.
• Google Ads segmentation: assigns each product to a segment (Hero, Sidekick, Villain, Zombie) based on its performance in Google Ads and writes it to the selected element.
• Google Analytics segmentation: assigns each product to a segment (Hero, Sidekick, Non-performing, Zombie) based on traffic and conversion rate from Google Analytics and writes it to the selected element.

You can also pass a document from the Mergado Contexts extension to the AI rules, for example your online store’s tone of voice (see Drawing data from documents).

Mergado Pilot is only available in Canvas (the new interface) — you will not find it in the classic interface.

Available rules by output format#

The availability of each type depends on the output format of your project. An unsupported type will appear in the list but cannot be selected.

Activation and access#

When adding a new rule, select Mergado Pilot as the rule type. A list of available Pilot rule types will appear — select the one you want to use and continue through the setup wizard.

Important: AI processing takes longer#

AI text processing runs asynchronously — in the background, outside the regular feed regeneration process. This means that changes may not be available after the first feed regeneration. The AI is still finishing processing the products.

Wait for Mergado to regenerate the feed again (or regenerate it manually) — only then will the AI results appear in the output. For large catalogues, you may need to regenerate the feed multiple times.

Important: Mergado Pilot modifies data only in the feed. Products in your online store remain unchanged.

Drawing data from documents (Mergado Contexts)#

For AI rules (Element enhancer, Q&A, Product highlight, and Custom prompt), you can pass the content of a document from the Mergado Contexts extension to the AI, for example your online store’s tone of voice or copywriting guidelines. The AI uses it as context when processing each product.

• In the Data and context step, turn on the Use data from documents toggle and select the document in the Document field.
• Pilot uses the first 5,000 characters of the document and ignores the rest.
• If you do not have the Mergado Contexts extension active, the step shows an Activate Mergado Contexts button instead of the toggle, which opens Mergado Store.

When you edit the document in Mergado Contexts, the change is applied to Pilot automatically within a few hours and the rule’s products are processed again. If you do not want to wait, click Refresh document in the rule edit view. Then regenerate the feed.

Rule: Title and description improvement#

Typical use cases:

• Optimising product titles for Google Shopping, Heureka, or Meta
• Rewriting descriptions for a specific market or language
• Adding missing information to descriptions based on feed data
• Unifying the tone of communication across your product catalogue

How the rule works#

The rule iterates through products from the selected queries and sends each product’s feed data along with your instruction to the AI. The AI returns the modified text, which Mergado writes to the target element (Title or Description) — just like any other rule.

Results depend on the quality of the data in your feed and how precise your instruction is.

Creating the rule#

The rule is created using a 5-step wizard.

Step 1 – Queries#

Choose which products the rule should apply to:

• All products — the rule updates the element (title or description) for all products in the project.
• Custom query — the rule applies only to products from the selected product query. You can select an existing query or create a new one directly in this step.

Step 2 – Element#

Select whether you want to update the product title or product description:

• Product title — a short title optimised for advertising. Mergado Pilot will update the element containing the product title (e.g. PRODUCTNAME or g:title).
• Product description — a longer text describing the product. Mergado Pilot will update the element containing the product description (e.g. DESCRIPTION or g:description).

A single rule run updates only one element at a time. To update both the title and description, create two separate rules.

Step 3 – Prompt#

Enter the instruction that tells the AI how to modify the text. There are two approaches:

a. Preset style:

Choose one of three styles — Mergado Pilot uses a prepared system prompt optimised for that tone:

• Factual and informative — precise, structured texts. Suitable for technical products.
• Friendly and clear — natural language for everyday customers. Suitable for fashion, food, gifts.
• Energetic and compelling — emotional, sales-driven tone. Suitable for promotions, new arrivals, premium goods.

System prompts are also tailored to the target output format of your project (Google Shopping, Heureka, Meta, and others). For example, if you are working on a Google Shopping output, the AI automatically takes into account Google Merchant Center recommendations and guidelines — without you having to specify them explicitly.

b. Custom prompt:

Write any instruction for the AI. For example:

• “Rewrite the product title in the format: Brand + Name + Colour + Material. Maximum 150 characters. Use only product data.”
• “Add key product features suitable for Google Shopping to the description. Use only data from the feed.”

Strict mode:

This step also lets you enable strict mode. When enabled, the AI does not add information that is not directly present in the feed data. Outputs are more accurate but may be shorter. When disabled, the AI can naturally expand and enrich the text.

When to use strict mode: Enable it when you want to make sure the AI does not invent parameters, colours, or attributes that are not in the feed. Especially useful for technical products or any catalogue where data accuracy matters.

Step 4: Data and context#

Optionally, select a document from Mergado Contexts for the AI to use as context (see Drawing data from documents).

Step 5: Settings#

Enter the rule name that identifies it in the project’s rule list.

Tip: Before applying the rule to your entire catalogue, test it on a smaller product query first. Review the results in the product preview, then expand the query to the full catalogue.

Examples#

Example 1 — Optimising titles for Google Shopping#

A clothing online store wants to add brand, colour, and material to product titles to meet Google Merchant Center recommendations.

1. Create a new rule → type Mergado Pilot → Element enhancer.
2. Queries: All products (or a query for a specific category).
3. Element: Product title.
4. Prompt: Custom — “Rewrite the product title in the format: Brand + Name + Colour + Material. Maximum 150 characters. Use only product data.”
5. Strict mode: On.
6. Save the rule and apply it. Results will appear in the feed after regeneration.

Example 2 — Friendly description for Heureka#

An online store wants product descriptions on Heureka to be readable and easy to understand for everyday customers — not just a technical list of specifications.

1. Create a new rule → type Mergado Pilot → Element enhancer.
2. Queries: All products (or a query for a specific category).
3. Element: Product description.
4. Prompt: Preset style — Friendly and clear.
5. Strict mode: Off — the AI can naturally enrich the text.
6. Save the rule and apply it. Results will appear in the feed after regeneration.

Rule: Q&A (question and answer)#

The rule automatically generates the content of the g:question_and_answer element for Google Shopping. The element contains question/answer pairs that can help customers make a purchase decision.

The data source is the product feed, primarily the product description. Optionally, the AI can also draw on customer reviews (see Using data from reviews). Processing runs using an LLM on Mergado’s servers — data does not leave its infrastructure.

Creating the rule#

The Q&A rule has a simple setup with just three steps:

1. Queries: choose which products the rule should apply to (all products, or a custom query).
2. Data and context: decide whether the AI should also draw on customer reviews or on a document from Mergado Contexts.
3. Settings: give the rule a name.

The number of generated question/answer pairs is determined automatically by the rule — it cannot be set manually.

Using data from reviews#

In the Data and context step, you can turn on the Use data from reviews toggle. The AI will then take into account facts from your online store’s customer reviews in addition to the feed data. It never quotes the reviews or mentions their source. The Product highlight rule has the same step.

You can only turn the toggle on if:

• you have the Mergado Depots extension enabled,
• you have an active connection to Heureka or Zboží.cz in Depots (one of them is enough).

If anything is missing, the toggle is inactive and the AI works with feed data only. The step shows a message explaining what to add, along with a button that takes you to Mergado Depots.

Example#

An electronics online store wants to answer common customer questions directly in Google Shopping before purchase (e.g. “Does the headset have wireless volume control?”). Simply create a Q&A rule for a selected product category — Pilot will automatically generate the corresponding question/answer pairs from the product description.

Rule: Product highlight#

The rule automatically generates the content of the g:product_highlight element for Google Shopping — the element records key product features and can appear multiple times for a single product.

The data source is the product description and, optionally, customer reviews (see Using data from reviews). Processing runs using an LLM on Mergado’s servers — data does not leave its infrastructure.

⚠️ This rule is currently only available for the Google Shopping format — for ChatGPT (OpenAI Shopping), the corresponding element does not yet exist.

Creating the rule#

Like Q&A, the Product highlight rule has just three steps:

1. Queries: choose which products the rule should apply to.
2. Data and context: decide whether the AI should also draw on customer reviews or on a document from Mergado Contexts.
3. Settings: give the rule a name.

Example#

A fashion online store wants to highlight key product features in Google Shopping (e.g. “Waterproof material”, “Handmade”) without having to write them manually for each product. The Product highlight rule generates them automatically from the product description.

Rule: Custom prompt#

A rule for advanced users who want full control over what the AI does. You choose what data the AI receives, which element it writes the result to, and what it should do with the data. Unlike the Title and description improvement rule, it does not use any preset system prompt: the AI follows only your instruction.

Typical use cases:

• Translating an element into another language
• Extracting specific information from the description into a separate element (for example, size into g:size)
• Modifying elements other than the title and description

Creating the rule#

The rule is created using a 6-step wizard:

1. Queries: select all products, or a custom product query.
2. Source data: select one or more elements to send to the AI. The more data you send, the more tokens are consumed and the slower the processing. Send only what the AI really needs for the result.
3. Target element: select the element the AI will write the result to. It can also be an element you selected as a source; the AI will then overwrite it. Nested elements cannot be selected yet.
4. Prompt: write exactly what the AI should do with the data.
5. Data and context: optionally, select a document from Mergado Contexts (see Drawing data from documents).
6. Settings: give the rule a name.

The rule has no strict mode or preset styles. If you do not want the AI to add information that is not in the data, say so directly in the prompt.

Examples#

Example 1: Translating the description#

1. Source data: g:description
2. Target element: g:description
3. Prompt: “Translate the product description into German.”

Example 2: Extracting the size from the description#

1. Source data: g:description
2. Target element: g:size
3. Prompt: “Find the product size in the description and return only the value itself, with no additional text.”

Rule: Google Ads segmentation#

The rule assigns products to segments based on their performance in Google Ads and writes the segment to the selected element (the default is custom_label_0). In Google Ads, you can then group products by element value and manage the budget accordingly (for example, separating top-performing products into a dedicated campaign). This is not an AI rule: the segment is calculated from real data on clicks, costs, and revenue.

Segments#

The value written to the element is always lowercase English: hero, sidekick, villain, zombie, other, or nodata. Products move between segments automatically based on current results.

What the rule needs#

The rule draws data from Google Ads via Mergado Depots, so you need:

• the Mergado Depots extension enabled,
• an active Google Ads source in Depots, connected via Mergado Keychain (see How to work with Mergado Depots).

If anything is missing, the rule form locks and shows a message explaining what to add. Once the connection is active, the rule displays the information Google Ads connection is active: Mergado Depots fetch data from Google Ads once a day, and Pilot recalculates segments after each nightly update. You will also see the time of the last data fetch. If no data has arrived yet, the message First data from Mergado Depots has not arrived yet — this usually takes up to 24 hours after connecting will be displayed.

Creating the rule#

The rule is set up in a wizard where you configure the following steps in order:

1. Queries: select all products, or a custom product query.
2. Element: choose the element the segment will be written to. The default is custom_label_0. If you already use it for something else, choose a different one to avoid the rule overwriting it.
3. Segmentation: set the time period and the values used to sort products:
   • Period length: 1 day, 7, 30 (default), or 90 days. A small online store with limited data should choose a longer period; a large online store can use a shorter one.
   • Minimum click count (CLICKS_MIN, default 50): the minimum number of clicks before the rule considers a product’s data sufficient.
   • Target return (TROAS, default 1000 %): the return above which a product is considered profitable.
4. Settings: give the rule a name.

Segmentation preview#

At the top of the rule edit view is the Segmentation preview panel. It shows a chart of product distribution and a legend with the count and percentage of products in each segment (Hero, Sidekick, Villain, Zombie, Other, No data; segments with no products are shown as zero). The Show products by segment link takes you to the product list.

The Recalculate preview button calculates the impact of unsaved values (period length, CLICKS_MIN, TROAS) without saving the rule. It is inactive until at least one value differs from the saved one, and shows the change relative to the saved settings for each segment. The preview is indicative only: the actual distribution changes only after saving the rule and regenerating the project.

The panel may show the following messages:

• Segmentation preview will be available after the first regeneration by Mergado. Shown after saving a new rule or changing its settings. Renaming the rule alone does not trigger this message.
• Products without a segment (waiting for regeneration): N. Regeneration is in progress and some products do not yet have a segment.
• No product has Google Ads data yet. Mergado Depots fetch data once a day — wait for the next fetch.

Example#

An online store wants to shift budget in Google Ads away from underperforming products towards top performers. They create a Google Ads segmentation rule for all products, keep the 30-day period, and adjust TROAS to match their target return. They check the preview to see how many products land in the Hero and Villain segments, save the rule, and after the feed is regenerated, group products in Google Ads by the value in custom_label_0.

Rule: Google Analytics segmentation#

The rule assigns products to segments based on traffic and conversion rate from Google Analytics and writes the segment to the selected element. In Google Ads, you can then group products by element value and prepare campaigns for each group, for example separating products that bring traffic but do not sell. This is not an AI rule: the segment is calculated from real data on visits and orders.

Unlike Google Ads segmentation, it works with visitor behaviour on your online store, not with advertising performance.

Segments#

The value written to the element is always lowercase English: hero, sidekick, non-performing, zombie, other, or nodata. Products move between segments automatically based on current data.

What the rule needs#

The rule draws data from Google Analytics via Mergado Depots, so you need:

• the Mergado Depots extension enabled,
• an active Google Analytics source in Depots, connected via Mergado Keychain (see How to work with Mergado Depots).

Mergado Depots fetch data from Google Analytics once a day.

Creating the rule#

1. Queries: select all products, or a custom product query.
2. Element: choose the element the segment will be written to. The default is custom_label_0. If you also use Google Ads segmentation, choose a different element for each rule, otherwise the values will overwrite each other.
3. Segmentation: set the time period and the values used to sort products:
   • Period length: 1 day, 7, 30 (default), or 90 days.
   • Target conversion rate (default 2 %): the conversion rate you consider good.
   • Minimum visit count (default 50): the minimum number of visits before the rule considers a product’s data sufficient.
4. Settings: give the rule a name.

As with Google Ads segmentation, the Segmentation preview panel is at the top of the rule edit view. Use the Recalculate preview button to see how the segment distribution would change after adjusting the values, without saving the rule.

Example#

An online store wants to find out which products get many visits on the website but do not sell. They create a Google Analytics segmentation rule for all products, write the segment to custom_label_1, and keep the default values. After the feed is regenerated, the Non-performing segment shows products whose price, availability, or description are worth checking, and they can adjust the budget for them in Google Ads.

FAQ#

What is the difference between Mergado Pilot and the Rewrite rule?#

The Rewrite rule writes exactly the text or variable you specify — the result is always the same for all products. Mergado Pilot sends each product’s data to the AI, which generates unique content for each product based on your instruction (or automatically for Q&A and Product highlight). Use it when you need content tailored to individual products.

How many products does the rule process at once?#

The rule processes all products from the selected query sequentially. Processing a large catalogue takes longer. The AI generates content for each product individually, and results are written to the feed asynchronously once complete. You may therefore need to regenerate the feed several times.

Can I create a separate rule for the title and the description in one project?#

Yes. Create two separate Title and description improvement rules, each for a different element. One will update the product title, the other the product description.

What if the AI output does not meet my expectations?#

For the Title and description improvement rule, edit the prompt or toggle strict mode and run the rule application again. For the Custom prompt rule, edit the prompt or the source data. For Q&A and Product highlight rules, custom settings cannot yet be adjusted. We recommend testing on a smaller product query first.

Does Mergado Pilot affect the order of other rules?#

Yes. AI rules behave like standard rules, and their position in the rule order affects what data the AI receives as input and how subsequent rules can modify it. If you want the AI to work with values already modified by other rules, place Mergado Pilot after them.

How can I check what the AI rule changed for a specific product?#

Use the Rule trace feature — it shows how the element value changes after each rule is applied, including the AI rule.

Can I enable and disable the rule without deleting it?#

Yes. The rule can be paused just like any other rule type. Products will remain unchanged until you activate the rule again.

What is strict mode and which rule has it?#

Strict mode is available for the Title and description improvement rule. It tells the AI to rely only on the product data in the feed and not to add information that is not there. Enable it when accuracy matters — for example, for technical products where you do not want the AI to invent a colour or specification.

Why does Q&A or Product highlight not appear for my project?#

These rules are only available for the Google Shopping format. For an unsupported format, the rule will appear in the list as unavailable.

Can I set a custom prompt or style for Q&A or Product highlight?#

Not yet. Unlike the Title and description improvement rule, these generate content automatically, without the option to set a custom prompt or choose a style. If you need your own instruction, use the Custom prompt rule.

Where do the Q&A and Product highlight rules get their data from?#

From the product feed, primarily from the product description. If you turn on Use data from reviews in the Data and context step, the AI also takes into account facts from customer reviews from Heureka or Zboží.cz connected via Mergado Depots. Optionally, you can also add a document from Mergado Contexts.

What is the difference between Title and description improvement and the Custom prompt rule?#

Title and description improvement modifies only the title or description and uses prepared system prompts that know the rules of advertising channels. Custom prompt has no system prompt: you choose the source elements, the target element, and the instruction yourself.

Is the content sent to the AI protected?#

Product data (title, description, and any parameters) is sent to an AI model for the purpose of generating content. If you use a document from Mergado Contexts in the rule, its content is also sent to the AI. The data is not stored for model training. For Q&A and Product highlight rules, processing runs on Mergado’s servers, so data does not leave its infrastructure. The system prompt that governs the AI’s basic behaviour is managed by Mergado and cannot be edited directly by the user. The exception is the Custom prompt rule, which does not use a system prompt.

How often are segments updated?#

Mergado Depots fetch data from Google Ads once a day. Pilot recalculates segments after each nightly update, and the results appear in the feed after the project is regenerated.

Why do products show the No data segment?#

Mergado Depots do not yet have Google Ads data for this product. It usually takes up to 24 hours after connecting, or the product has no Google Ads data for the selected period.

Why is the Google Ads segmentation rule form locked?#

The rule requires the Mergado Depots extension to be enabled with an active Google Ads source (connected via Mergado Keychain). The message in the form tells you what is missing.

Does the segment change immediately after saving the rule?#

No. Segments are written during project regeneration. Until then, the Segmentation preview panel shows a message that the preview will be available after regeneration.

Which element is the segment written to?#

The default is custom_label_0. In the Element step, you can choose a different one — for example, if custom_label_0 is already used for another value.

What is the difference between Google Ads segmentation and Google Analytics segmentation?#

Google Ads segmentation sorts products by advertising performance (clicks, costs, return). Google Analytics segmentation sorts them by visitor behaviour on your online store (visits, conversion rate). You can use both rules at the same time, but each must write to a different element.

Can I edit the rule while products are still being processed?#

Yes. If you change anything other than the rule name or the React to changes in elements toggle, Pilot asks you to confirm. After saving, it discards the rule’s existing results and processes the products again according to the new settings during the next feed regeneration.