Answer the second question: a playbook for AI follow-up questions on CPG product pages

A broad shopping search used to return ten blue links and let the shopper do the narrowing. Google's AI answers now do some of that narrowing themselves, by asking questions back.

Natasha Post described this in Search Engine Land on November 2. She searched for birthday gifts for a 10-year-old who loves Legos, Marvel, fishing and video games. Below the AI Overview, Google asked follow-ups before she had visited a single site:

• What is your approximate budget?
• Does he play video games on a specific console?
• Would you prefer something hands-on to build or an outdoor gear item?

Each answer produced a narrower set of results. Her point for marketers was that clicks can now come from almost any stage of the buying journey, and content has to anticipate the next question, not just the first one.

That applies to packaged goods at least as much as to toys. A shopper asking for "a good snack for my kid's lunchbox" is one or two follow-ups away from "nut-free," "under $5" or "no added sugar." This playbook is how we would prepare a CPG brand for those questions.

Why the second question matters

The follow-up is a filter. When the assistant asks about budget and the shopper says "under $10," every product without a clear price signal becomes harder to include. When it asks about diet and the shopper says "gluten-free," a product that only mentions gluten in an image of the back of the pack may not make the cut.

The behavior data points the same way. Fractl's clickstream panel, published with Search Engine Land, shows people staying in search longer rather than giving up.

20242026

Show the data

Behavior	2024	2026
Continued searching (re-query)	54%	63%
Zero-click bounce	43%	34%

Source: Fractl and Datos clickstream panel, via Search Engine Land. Equal-weighted per keyword across roughly 11,100 keywords; directional, not population-level rates.

More refining means more steps where a product can be added or dropped. Brands used to optimize for the first query. The follow-ups are where a lot of the choosing now happens.

Step 1: Pick three shopper paths per product line

Start with the queries real shoppers use, not your brand terms. For a CPG product line, three paths are usually enough to learn a lot:

1. The need path. The problem the product solves, in the shopper's words: "something to help my toddler sleep," "dish soap that is gentle on hands."
2. The occasion path. The moment: "snacks for a road trip," "what to bring to a potluck," "stocking stuffers for coworkers."
3. The comparison path. The category head-to-head: "best plant-based protein powder," "is store brand oatmeal as good as name brand."

Write each starting query as a full sentence. AI answers handle long, specific questions, and our post on how AI Mode reads long queries explains why product copy should be ready for them.

Step 2: Walk each path by hand

Run each query in Google and answer the follow-up questions one at a time, as the shopper would. Then start over and answer differently: a lower budget, a different diet, a different age. Post's advice is to do this yourself rather than relying only on tools, because the follow-ups change with the person and over time.

Record four things for every run:

• The follow-up questions asked, word for word.
• The answer you gave.
• Which brands and products appeared after that answer.
• Whether your product appeared and, if it did, what the answer said about it.

Do this signed out or in a clean browser profile where you can, and note the date. Google's guidance on AI features says there are no special optimizations needed to appear in AI Overviews or AI Mode, and that the same foundational SEO practices apply. The follow-ups you see are a useful sample, but personal context can shift them.

Step 3: Log which questions repeat

After a dozen runs you will see patterns. Post suggests documenting three things: which questions repeat, which answers bring up your company, and which answers should bring it up but do not. That third list is the work queue.

Here is the kind of log we keep. The questions and gaps below are an illustrative example for a fictional snack bar brand, not observed results.

Follow-up question (example)	Shopper answer (example)	Product shown?	Where the answer should live
Is this for a child or an adult?	Child, age 6	No	Product page: "made for kids' lunchboxes" in plain text
Any allergies to avoid?	Nuts	No	Product page, retailer listing and feed: "nut-free, made in a nut-free facility" if true
What is your budget?	Under $6 a box	Yes	Retailer listing price and count, per-bar price in copy
Do you prefer low sugar?	Yes	No	Nutrition panel as HTML text, sugar grams per bar stated
How many do you need?	A week of lunches	Yes	Count per box in the title and first bullet

The last column is the useful one. For every follow-up where the product should qualify but does not appear, write down the plain sentence that would answer the question, and the places it needs to appear.

Step 4: Put the answers where assistants read

Follow-up answers are usually small facts: age, allergen, count, price tier, sugar, size, occasion. The work is putting each one in plain, matching text in several places.

On your product page. State the facts as text, not only in images. A short "Good to know" block works: who it is for, what it is free of, how many per pack, how it is used. Nutrition and ingredient details belong in an HTML table, which our tip on HTML tables covers.

On retailer listings. Most product follow-ups lead to a retailer. The title, bullets and attributes should carry the same facts with the same wording as your site. Mismatches give an assistant a reason to hedge or skip, and our tip on matching product facts has the checklist.

In an FAQ. A short FAQ on the product page can answer the follow-ups you logged in the shopper's own words: "Is it safe for school lunches?" "How much sugar per bar?" Keep each answer to one or two plain sentences.

In your feed. Google's shopping data comes partly from Merchant Center, and its product data specification lists the attributes it accepts, such as size and multipack count. Whatever you submit there should match the page.

Do not stuff the page with every possible question. Answer the ones that repeat, honestly. If the product is not nut-free, do not imply it, and make the actual allergen statement clear. An assistant that gets a fact wrong is worse than one that leaves you out, and our tip on correcting wrong AI answers covers what to do when that happens.

Step 5: Re-run monthly and watch the gaps close

Follow-up questions are not fixed. They depend on the person, the query and Google's current models. Keep your log, re-run the same paths monthly on the same day, and compare.

The measure that matters is simple: of the follow-up answers where your product should qualify, in how many does it now appear? That number moving from, say, one in five to three in five is real progress, even if the first answer never mentions you. A spreadsheet is enough to start. Teams that want to repeat shopper-style prompts across several assistants can use a dedicated tracker; SearchPanel, built by XENA Intelligence, is one example.

A note on what not to chase

It is tempting to read the follow-up trend as a reason to publish a page for every possible variation of a question. We would not. Post's own conclusion is about knowing your customer well enough to anticipate the next question, not about volume. For CPG, that means a short list of facts per product, stated consistently everywhere. The quiet click data also suggests that when shoppers do click, they are often checking whether the AI was right. The best follow-up answer is the one that matches what they find when they check.

The broad query still matters. But the brand that answers the second and third questions plainly is the one that stays in the set when the shopper finally decides.