The 2026 SERP is no longer a list of ten blue links — it’s three distinct territories to conquer
When a user types a question on Google in 2026, they no longer receive a SERP. They receive a patchwork: an AI Overview at the top (present on 47% of informational queries per Semrush May 2026 measurements), a People Also Ask block expanding dynamically (up to 8-12 stacked questions on long queries), a classic Featured Snippet (still there, still strategic), then only the organic list. On Bing, Copilot occupies half the screen. On ChatGPT and Perplexity, there’s no SERP anymore — there’s a synthesised answer with a few clickable citations.
For an e-commerce store, the stake is double. First, these three territories (Featured Snippets, PAA, AI Overviews) concentrate 60 to 75% of search traffic in 2026. Second, they aren’t won with the same content: a Featured Snippet likes a short definition, a PAA likes a 40-60 word answer, an AI Overview likes structured content with named entities and factual data. Content that conquers all three is content written for all three — and that doesn’t happen by itself.
This article details the mechanics of each territory, the content format winning on each, and how to automate this production by AI on PrestaShop and WooCommerce.
Territory 1 — Featured Snippets: the position 0 snippet
The Featured Snippet has existed since 2014. It occupies the top of the SERP with an answer extracted from a web page. Three main formats:
- Paragraph: 40-58 words answering a question (“what is”).
- List: ordered or not, 4 to 8 items (“steps to”, “how to do”).
- Table: structured comparison (“difference between A and B”, “average price”).
What makes a Featured Snippet win in 2026
- Question explicitly posed as H2 or H3 heading, immediately followed by the answer in fewer than 60 words.
- Answer above the fold of the page (Google scans the first half more thoroughly).
- Structured data Schema.org FAQPage or HowTo aiding extraction.
- Topical authority of the site — Google doesn’t give a Featured Snippet to a site that isn’t already well-ranked on the term.
For an e-commerce product page
The Featured Snippet on a commercial query (“best latex mattress”, “difference between PrestaShop and WooCommerce”) is highly coveted. The strategy that works:
- Identify long-tail queries with low competition.
- Write an article or enrich a product page with H2 = the exact question of the query, followed by a short, precise answer.
- Add
FAQPageSchema.org markup on the question/answer.
The effect: a well-captured Featured Snippet multiplies classic position 1 CTR by 4 to 8, and survives even partially above an AI Overview (Google has let Featured Snippets and AI Overviews coexist since 2024).
Territory 2 — People Also Ask: the expanding block
The PAA block presents 3 to 5 “related” questions under the SERP. Each click on a question displays an answer extracted from a page, and adds 2 to 3 new questions to the block — which can expand to 12-20 questions in an engaged user session.
PAA opportunity is massive and under-exploited
Three characteristics make PAA strategic:
- PAA is more accessible than Featured Snippet: it accommodates several different pages (3 to 5+), not just one. Competition is mechanically less violent.
- A question appears in 50 to 200 different parent queries. Capturing a PAA question delivers traffic across the entire tree of associated questions.
- PAA citation gives brand visibility even when the user doesn’t click — the extract with the source reference reinforces awareness.
The content format capturing PAAs
A PAA question likes an answer of 40 to 60 words, placed right after the question. Not a long contextualising paragraph before. The Google crawler extracts the answer linearly after the H3 corresponding to the question.
Type structure that wins:
<h3>How much does a PrestaShop module cost?</h3>
<p>A PrestaShop module costs between £30 and £500 in
perpetual licence, depending on complexity. SEO and
marketing modules are generally £50-150, ERP/B2B modules
£200-500. Free modules exist but have limited support.</p>53 words. Direct question in H3. Self-contained answer (readable out of context). Concrete numeric data. That’s the profile maximising PAA capture probability.
Generating the right questions by AI
Manual identification of PAA questions goes through AlsoAsked.com or Semrush — excellent tools but requiring human sorting. The 2026 alternative: generate questions by AI from product content, then validate they correspond to real PAAs.
The DfAiPAA module for PrestaShop implements this workflow: for each product page or category, it generates via GPT-4o-mini a list of 8 to 15 likely questions, crosses them with a SERP API (DataForSEO or similar) to verify they appear in Google PAA, and proposes answer content. AI cost: €0.03 to €0.08 per enriched page. The module also handles automatic Schema.org FAQPage markup on each injected Q/A.
Territory 3 — AI Overviews and AI agents
Google AI Overviews (launched in May 2024 as Search Generative Experience then renamed in 2024) generate a synthetic answer at the top of SERPs on complex or informational queries. This answer typically cites 3 to 8 sources. Being cited in an AI Overview is the modern equivalent of a Featured Snippet — except instead of capturing the click, you capture brand visibility AND (in 35 to 45% of cases) a click to the source.
In parallel, ChatGPT Search, Perplexity, Claude (via shopping mode), and Bing Copilot build their own answers with their own attribution rules.
What AI agents look for in a page
Three signals dominate in 2026:
- Explicit named entities — Schema.org Product, Organization, FAQPage, HowTo with their complete properties.
- Factual numeric data — prices, dimensions, dates, statistics, comparisons.
- Clean semantic structure — single H1, H2 per section, H3 for questions, self-supporting paragraphs.
The llms.txt file has become a de facto standard for indexing a site to LLMs. It provides an AI agent with a map of indexable content, equivalent to robots.txt but for language models. On PrestaShop, automatically generating it from the catalogue is an immediate AI visibility gain.
The decisive role of product FAQs
A well-structured product FAQ — 5 to 10 questions/answers on each page — is the content simultaneously serving all three territories:
- For Featured Snippets: each Q/A is candidate for a snippet on its question.
- For PAAs: questions often match Google PAAs.
- For AI Overviews: the FAQPage Schema.org structure is extremely readable by LLMs.
The empirical observation: on stores with a well-filled AI FAQ (10+ questions per page), the share of AI-attributed traffic (from ChatGPT, Perplexity, Claude) reaches 4 to 7% in 2026 — vs 0.2 to 0.8% on stores without FAQ. It’s a channel that didn’t exist in 2023 and that today weighs as much as Bing organic.
Automating production: three complementary modules
1. AI product FAQ generation
On PrestaShop, the DataFirefly Product FAQ AI module generates for each page 5 to 10 contextualised questions/answers (from title, description, attributes, brand). Multi-language (FR, EN, ES, DE), automatic Schema.org FAQPage markup, batch generation. AI cost: €0.02 to €0.05 per page.
On WooCommerce, the equivalent is DataFirefly Product FAQ AI WooCommerce, with the same features adapted to the WC data model (custom fields, taxonomies, variations).
2. Contextualised PAA generation
The DfAiPAA module for PrestaShop goes further: it generates questions specifically aligned with Google PAAs detected by SERP crawl. Difference with classic FAQ: FAQ targets likely questions, PAA targets SERP-verified questions. Both are complementary.
3. llms.txt file
The llms.txt module for PrestaShop generates and maintains up-to-date the /llms.txt file at the site root, with the list of categories, product pages, blog articles, and their canonical URLs. It’s become a signal read by ChatGPT, Perplexity, Claude, and progressively by Google crawlers. Zero API cost (the file is static).
The WooCommerce counterpart implements the same logic with WP/WC structure.
The complete 2026 pipeline
For a store wanting to occupy all three territories, the recommended 2026 pipeline:
- Initial audit — Identify the 50 to 100 priority queries (top categories + top pages + long-tail queries identified in Search Console). For each: Featured Snippet/PAA/AI Overview presence? Competitors occupying them?
- Product FAQ generation — AI module deployed across the entire catalogue. 5 to 10 Q/A per page, Schema markup.
- Enriched PAA generation — For the top 20% pages generating 80% of traffic, enrich with specific PAAs crawled from Google.
- Targeted editorial articles — For top-funnel queries (“how to choose”, “difference between”): dedicated blog articles with Q/A structure, FAQPage markup, and internal linking to product pages.
- llms.txt file deployed — Daily update, listed in
robots.txtviaSitemap:. - Measurement — Search Console for PAA/Featured Snippets, UTM parameter or referrer for AI traffic (ChatGPT has sent its referrer since 2024).
Typical ROI on a mid-market PrestaShop or WooCommerce store: +18 to +35% organic search traffic in 6 to 9 months, of which 4 to 8 points specifically attributable to AI traffic.
Link with internal linking and global AEO
FAQ and PAA production only makes sense if the content is correctly linked. An orphan FAQ receives little authority; an FAQ linked from 8 to 15 catalogue entities ranks in a few weeks. It’s the articulation with AI semantic internal linking that transforms content production into real SEO performance.
Similarly, PAA/FAQ work fits into a global 2026 AEO approach that also covers Schema.org Product 2026 structured data and consumption fluidity by AI shopping agents.
The pitfalls that kill ROI
1. Mass generation without quality control
Generating 10,000 Q/A by AI in one night is technically possible. It’s also the best way to flood your site with mediocre content that Google identifies as “bulk AI-generated” and triggers a Helpful Content Update penalty. The rule: validate sampling 10-15% of generated Q/As, adjust the system prompt if needed, and purge Q/As that add no information.
2. Inconsistent FAQPage Schema
The FAQPage Schema.org markup must describe Q/As actually visible on the page. If markup lists 10 questions but visible content only displays 5, Google considers this structured data spam and deindexes the markup. Always sync visible and marked-up content.
3. Cannibalisation between pages and articles
If the same question appears with the same answer on 200 catalogue pages, Google doesn’t know which to rank. Deduplication is crucial: a generic question goes in a blog article; specific questions (product, size, colour) stay on the page.
4. Invented PAAs
Generating by AI “plausible” questions corresponding to no real Google PAA is dead content. Always cross-check with a SERP API (DataForSEO, Serper, ValueSerp) to validate the question has searches behind it.
FAQ
How many questions per product page is the optimum?
The 2026 sweet spot is 7 to 12 questions per page. Fewer than 5: under-exploits the territory. More than 15: Google considers the page “FAQ-loaded” and devalues. Very technical pages (electronics, IT) can go to 15-20 if quality is there.
Should they be displayed fully expanded or as accordion?
Accordion is OK for UX but content must remain in HTML (not loaded in JavaScript). Google and LLMs read rendered HTML, not UI state. If content is only revealed on click via JavaScript fetch, it’s not indexed. The rule: content in static HTML, JS for visual toggle only.
Does AI generated FAQ pose a problem with Google?
Not in itself. Google has specified since 2023 that quality prevails over origin. An AI-generated FAQ, factual, useful, without hallucinations, contextualised to the page → OK. A mass-generated FAQ, generic, repeating from page to page → Helpful Content Update.
Is AI traffic (ChatGPT, Perplexity) measurable?
Yes since 2024. ChatGPT sends the referrer chatgpt.com, Perplexity perplexity.ai, Claude claude.ai. In GA4, create a Custom segment based on these referrers for precise reading. Cutting-edge stores today measure 3 to 8% of their search traffic from these sources.
How to optimise specifically for Bing Copilot?
Bing Copilot consumes largely the same signals as Google (Schema, structure, authority). Well-done Google optimisation covers 80% of Bing. The specific 20%: submit the sitemap in Bing Webmaster Tools, verify indexing, and favour content with explicit citations (Bing values pages clearly citing their sources).
In summary
The 2026 SERP breaks down into three distinct territories — Featured Snippets, People Also Ask, AI Overviews and AI agents — calling for specific content formats but converging on a common base: structured Q/A, clean Schema.org FAQPage, and an up-to-date llms.txt file. Manual production of this content doesn’t scale; well-controlled AI generation largely does.
On PrestaShop, the recommended stack combines three complementary modules: Product FAQ AI for catalogue coverage, DfAiPAA for real Google PAA targeting, and the llms.txt module for LLM indexing. On WooCommerce, the FAQ AI module covers the same scope adapted to the WordPress ecosystem.
Typical 2026 ROI on a mid-market store: +18 to +35% organic search traffic in 6 to 9 months, of which 4 to 8 points specifically attributable to traffic from AI agents. AI setup payback (modules + inference cost) is around 60 to 90 days. Our AEO audit identifies priority targeting and calibrates production to avoid the pitfalls of mass generation.
