AI People Also Ask — Complete documentation (dfaipaa)

Overview

AI People Also Ask (technical slug: dfaipaa) captures the questions your customers actually ask Google — the “People Also Ask” blocks — generates the answers with the AI of your choice, and publishes a schema.org-marked FAQ on your product and category pages.

The module industrialises a complete four-step pipeline:

• Scraping — capture PAA questions on your target keywords via SerpApi, DataForSEO or manual entry.
• AI generation — write answers via Mistral, OpenAI or Anthropic, with configurable tone and brand voice.
• Editorial workflow — review, assign to products and categories, publish (manually or automatically).
• Publishing — accessible FAQ accordion on the storefront plus FAQPage JSON-LD for Google and generative engines.

Requirements

• PrestaShop 8.0.0 → 9.99.99
• PHP 8.1, 8.2 or 8.3
• MySQL 5.7 / MariaDB 10.4 or above
• An API key for at least one AI provider (Mistral, OpenAI or Anthropic)
• Optional: a SerpApi key or DataForSEO account to automate scraping
• Outbound HTTPS connections (cURL) allowed by your host

Installation

1. Download the dfaipaa.zip archive from your DataFirefly account.
2. In the PrestaShop back-office, go to Modules › Module Manager › Upload a module.
3. Drag and drop the ZIP, wait for confirmation, then click Install.
4. A new AI People Also Ask menu appears in the left column, with three tabs: Configuration, Keywords, Questions.

Installation creates 4 tables (prefix ps_dfaipaa_), sets the default configuration values and installs 4 admin tabs (one parent plus three children) with labels localised in FR, EN, ES, DE, IT and NL.

Configuration — Scraping

AI People Also Ask › Configuration tab, first section.

Getting a SerpApi key

Create an account on serpapi.com. The free plan gives 100 requests per month, roughly 100 scraped keywords. The key is in your dashboard under “Your Account”. The module queries the Google search endpoint and reads the related_questions block from the response.

Getting a DataForSEO account

Create an account on dataforseo.com. You receive a login / password pair to paste into the API key field as login:password (the module handles HTTP Basic authentication). DataForSEO bills per use, which suits larger volumes better. The module uses the SERP Google organic live advanced endpoint and extracts people_also_ask elements.

Location code mapping is built in for the following markets: FR, BE, CH, LU, CA, US, UK, IE, ES, PT, IT, DE, AT, NL, PL, BR and MX.

Manual entry mode

Select manual to disable all external calls. You then add questions yourself from the Questions tab; AI generation remains fully functional.

Configuration — Artificial intelligence

Second section of the Configuration tab.

Recommended models

• Mistral — mistral-large-latest for quality, mistral-small-latest to cut costs at high volume.
• OpenAI — gpt-4o-mini offers excellent value; gpt-4o for demanding technical catalogues.
• Anthropic — claude-sonnet-4-6 for nuanced, well-structured answers.

Constraints enforced on the model

The module builds a strict system prompt, independent of the provider: 60 to 120-word answers, simple HTML only (paragraphs, bold, italic, lists), no markdown, no heading tags, no scripts. The entity context (product or category name and description, truncated to 1200 characters) and the source keyword are injected to anchor the answer. The original Google snippet is supplied as reference with an explicit instruction to rephrase — never to copy.

Configuration — Display

Third section of the Configuration tab.

In tab mode, the module relies on PrestaShop’s native ProductExtraContent mechanism: the FAQ appears as a tab alongside “Description” and “Product details”, with no template override.

Editorial workflow

Step 1 — Add keywords

Keywords tab. Paste your list into the text area, one keyword per line, then submit. Duplicates are ignored automatically (adding is idempotent per keyword, language and shop).

Choose keywords that match purchase intent: “automatic coffee machine”, “best coffee beans”, “coffee maker maintenance”. Avoid pure brand queries, which rarely trigger PAA blocks.

Step 2 — Scrape

Two options:

• Scrape — individual button on each keyword row, handy for testing the configuration.
• Scrape all stale — processes in batches of 20 the keywords whose last capture exceeds the refresh interval.

Every captured question is stored with a uniqueness hash (question + language + shop): re-scraping a keyword never creates duplicates, it simply updates the last-capture date.

Step 3 — Generate answers

Questions tab. Filter by pending status, select questions with the checkboxes, then run the Generate bulk action. An individual button is also available on each row.

The entity context is built from the question’s first assignment. For better answers, assign the question to a product or category before generating: the AI will then have the entity’s name and description.

Step 4 — Review and assign

Click a question to open the edit form. You can:

• fix the HTML answer in the rich editor;
• assign the question to one or more products and categories (N-to-N relation);
• reorder assignments to control the accordion display order;
• reject an off-topic question (status rejected, kept in the database but never displayed).

Step 5 — Publish

Switch the status to published. The FAQ appears immediately on the storefront, along with its JSON-LD.

If Auto-publish is enabled in the configuration, steps 4 and 5 are merged: generation publishes directly. Convenient for a fully automated pipeline, best reserved for catalogues where human review is not critical.

Question statuses

Storefront display

The accordion is built on the native HTML details and summary elements, which guarantees:

• working keyboard navigation without JavaScript;
• content indexable by search engines even when collapsed;
• compatibility with every modern browser.

The first item is open by default. A lightweight CSS file is loaded and can be fully overridden from your child theme. All classes use the dfaipaa-faq prefix to avoid collisions.

JavaScript events

The front-end script emits two custom events you can wire into your analytics tool:

document.addEventListener('dfaipaa:open', function (e) {
  // e.detail.question, e.detail.index, e.detail.type, e.detail.entityId
  gtag('event', 'faq_open', { question: e.detail.question });
});

document.addEventListener('dfaipaa:close', function (e) {
  console.log('FAQ closed:', e.detail.question);
});

The views/js/front.js file also contains a SINGLE_OPEN constant (false by default): set it to true to allow only one panel open at a time.

Deep-linking

An anchor of the form #dfaipaa-q-123 automatically opens the matching question and scrolls the page to it. Handy for sharing a specific answer from an email or a support ticket.

FAQPage JSON-LD markup

On every product or category page load that has at least one published question, the module injects a JSON-LD block just before the document body closes (hook displayBeforeBodyClosingTag).

Emitted structure: a FAQPage node, a mainEntity array, and for each entry a Question node containing an acceptedAnswer of type Answer. The HTML content of answers is sanitised before emission: script and style tags along with event attributes are stripped.

Cron automation

A CLI script is provided to run the pipeline without manual intervention.

# Scrape stale keywords (20 max by default)
php modules/dfaipaa/cli/cron.php scrape --limit=20

# Generate AI answers for pending questions
php modules/dfaipaa/cli/cron.php generate --limit=10

# Chain scraping then generation
php modules/dfaipaa/cli/cron.php all --limit=20

Example crontab, nightly run at 3 am:

0 3 * * * cd /var/www/prestashop && php modules/dfaipaa/cli/cron.php all --limit=30 >> /var/log/dfaipaa.log 2>&1

Multilingual and multistore

Questions are indexed by id_lang and id_shop. In practice:

• the same keyword scraped in French and English produces two distinct sets of questions;
• answers are generated in the question’s language, the prompt applying an explicit language directive (fr, en, es, de, it, nl, pt, pl);
• in multistore, one shop’s questions and assignments never appear on another;
• display titles (tab, product, category) are stored as localised configuration.

Troubleshooting

Scraping returns no questions

• Check your quota at the provider: SerpApi silently cuts off beyond the free plan.
• Check language / country consistency: “fr” with “US” gives erratic results.
• Some keywords simply do not trigger a PAA block on Google. Test the query manually in a private browsing window.
• For DataForSEO, verify the login:password format of the API key field.

The AI returns markdown instead of HTML

Lower the temperature to 0.2, or switch to a more capable model. The prompt already enforces strict HTML rules, but the lightest models may partially ignore them.

The FAQ does not appear on the storefront

• Check that at least one question has the published status.
• Check that it is assigned to the entity being viewed (product or category).
• Check that display is enabled for that entity type in the configuration.
• Clear the Smarty cache from Advanced parameters › Performance.

JSON-LD does not appear in the source code

Make sure the “Emit JSON-LD” option is enabled and that your theme actually calls the displayBeforeBodyClosingTag hook. Some third-party themes omit it: in that case add {hook h='displayBeforeBodyClosingTag'} before the body closes in your layouts/layout-both-columns.tpl.

“Class not found” error after installation

The vendor/ folder was not extracted. Reinstall the module by uploading the ZIP through the back-office rather than extracting it manually.

Reading the operation logs

All operations (scraping, generation, publishing) are logged. To investigate:

SELECT * FROM ps_dfaipaa_log ORDER BY date_add DESC LIMIT 50;

Uninstallation

From Modules › Module Manager, click Uninstall. The operation drops the 4 ps_dfaipaa_* tables, the 4 admin tabs and all DFAIPAA_ configuration keys. Generated content is permanently lost: export your questions beforehand if you want to keep them.

Technical reference

• Technical slug: dfaipaa
• Namespace: DataFirefly Dfaipaa (PSR-4, bundled autoloader)
• Tables created: ps_dfaipaa_keyword, ps_dfaipaa_question, ps_dfaipaa_assignment, ps_dfaipaa_log
• Hooks used: displayHeader, displayProductExtraContent, displayFooterProduct, displayCategoryFooter, displayBeforeBodyClosingTag, actionFrontControllerSetMedia, actionAdminControllerSetMedia, actionProductUpdate, actionProductSave, actionCategoryUpdate, actionObjectProductDeleteAfter, actionObjectCategoryDeleteAfter
• Back-office tabs: AdminDfaipaa (parent), AdminDfaipaaConfig, AdminDfaipaaKeywords, AdminDfaipaaQuestions
• Configuration keys: DFAIPAA_SCRAPER_PROVIDER, DFAIPAA_SCRAPER_API_KEY, DFAIPAA_SCRAPER_LANG, DFAIPAA_SCRAPER_COUNTRY, DFAIPAA_SCRAPER_MAX_PER_KEYWORD, DFAIPAA_AI_PROVIDER, DFAIPAA_AI_MODEL, DFAIPAA_AI_API_KEY, DFAIPAA_AI_TEMPERATURE, DFAIPAA_AI_MAX_TOKENS, DFAIPAA_AI_TONE, DFAIPAA_AI_BRAND_VOICE, DFAIPAA_AUTO_PUBLISH, DFAIPAA_REFRESH_INTERVAL, DFAIPAA_PRODUCT_MODE, DFAIPAA_EMIT_JSONLD, DFAIPAA_TAB_TITLE, DFAIPAA_PRODUCT_TITLE, DFAIPAA_CATEGORY_TITLE
• CLI: modules/dfaipaa/cli/cron.php (commands scrape, generate, all)
• Front template: views/templates/hook/faq.tpl

GDPR compliance

The module neither collects nor stores any personal data: only keywords, questions, generated answers and technical operation logs are recorded. No cookie is set on the storefront. Calls to external APIs (scraping, AI) transmit only the keyword, the question and the product context — never customer data.

Support

For any technical question, contact the DataFirefly team at [email protected] or visit your customer area on datafirefly.com.