Your landing page can convert well and still be the wrong page for an AI Overview to cite. The mistake is trying to turn the whole sales page into an answer article, then wondering why Google quotes a plainer page instead.

I made that mistake for years. A page built to sell often makes an answer engine work too hard to find a sentence it can lift. The better setup is a two-page system: make the useful facts on the landing page extractable, build a companion page that answers the buyer’s question directly, and connect the two so a reader can move from answer to offer.

Why doesn’t the landing page get cited when it ranks?

Google does not require a special AI Overview optimization for eligibility. If a page is indexed and allowed in snippets, it can appear in AI Overviews or AI Mode; Google says there are no extra optimizations required for eligibility. That is easy to misread as a promise. Eligible is not citable.

I learned the difference watching a client landing page rank on page one while an Overview cited a forum thread with worse design and better specifics. The sales page had plenty to say about the offer but no clean passage that answered the question on its own. A primer on getting cited makes the same point: an extractable, self-contained statement is easier to quote out of context than polished copy that buries its specifics.

Ranking earns consideration, not a citation. One analysis of 863,000 keywords and 4 million Overview URLs found that 38% of cited pages also ranked in the top 10 for the same query, down from 76% in July 2025, while 31% of citations came from beyond position 100. Citation happens at passage level: a page is retrieved, then a passage has to answer clearly enough to stand alone. That is also the pattern Ten Speed describes in its client work.

The practical move is not another round of hero-copy revisions. Give the engine an answer it can use.

Claymation sales page shouting while a search engine lifts a plain answer card

Step 1: Which buyer questions does this page need to answer?

I start with the sales inbox, not a keyword tool. Pull the five to seven questions buyers ask before they convert: what it costs, what is included, how it compares with the alternative they mention, who it fits, how long it takes, and what can go wrong. Write them in the words buyers use. Those questions decide what belongs on the landing page and what needs room elsewhere.

Google can fan one query into several sub-queries and retrieve an answer for each. A broad page with no clean answers can lose to narrow passages that address those questions directly. The answer-block pattern is simple: a heading that sounds like the question, a first sentence that names the subject and answers it, then a few sentences of support. It is less glamorous than a content framework. It also gives the engine something to lift.

My working limit is five questions for one landing page. Beyond that, use the companion answer page. If you sell garage door installation in Austin, the landing page might cover price range, timing, warranty, and what affects cost. It does not need to explain how torsion springs work between the offer and the form. I cover that division in my passage-first guide to getting cited.

Do this before writing: review the last 90 days of sales calls, chat logs, and Search Console queries. Highlight repeated questions and assign each one to the landing page or the answer page. Do not make both pages fight to be the same article.

Step 2: What should you change on the landing page?

Start with what I call the extraction layer: the facts a buyer needs and an engine can read. Put the offer, price range, and core specifications in plain HTML text, not only in images, sliders, or a pricing graphic. Say you spend $20k a month on ads and display three tiers as PNGs. The page may look perfectly clear to a visitor, but its numbers are still harder to extract than a plain sentence in the page text.

I have watched Overviews cite a competitor with uglier design because its page stated the useful facts outright: installation included, a five-year parts warranty, service in Travis and Williamson counties. Your hero image may sell the work. That sentence can answer a question. This is why I tell clients to fix the myths that keep landing pages out of AI Overviews before they buy another tool.

Then add two or three short answer passages under question-style H2s. Keep the offer and form prominent; these blocks should support the sale, not shove it aside. Give each block the same shape:

  1. Ask the question in the heading.
  2. Answer it in the first sentence, naming the subject rather than writing “it” or “this.”
  3. Add a couple of sentences of context, including a concrete number when you have one.

An illustrative pricing block could read: “How much does garage door installation cost in Austin? Garage door installation in Austin costs $890 to $1,850 installed, with most standard 16-foot steel doors landing near $1,200. Price moves with insulation rating, opener type, and haul-away.” The figures are an example, not a price to paste onto a real page. The shape is the point: that answer makes sense without the paragraph above it.

If you add FAQPage or Product markup, match the visible copy. Do not treat markup as a substitute for it. A matched test of 1,885 pages found that adding JSON-LD moved AI Mode by 2.4% and ChatGPT by 2.2%, while AI Overviews fell 4.6%. I would keep schema for its other uses, not buy it as a citation tactic.

The rule: put the fact in readable page text first. Leave the rest of the landing page free to do its job.

Step 3: What goes on the companion answer page?

The answer page has one job the landing page cannot always do: hold still and answer in full sentences. I build it answer-first. The title names the buyer question. The first 60 words give a direct answer with the subject named plainly. A short comparison table follows, then the explanation a buyer needs to judge the choices.

For the garage-door example, the opening might compare an insulated 16-foot door at around $1,200 installed with an uninsulated option at $950. A three-column table can put price, insulation, and warranty side by side. Then use the rest of the page to explain when each option makes sense. Those example prices still need to be replaced with the business’s actual, current offer. The structure comes from the five page blocks built for AI answers that I keep bookmarked: direct answer, useful detail, no clever headline hiding the point.

Do not add a hero video, pop-up, or long introduction before the answer. A buyer came with a question. Make the first screen earn the next one.

Then connect that page to the sale. After the table, add a natural invitation to get a quote on the relevant landing page. Repeat the link at the end if it helps the reader who finishes the explanation. That is one destination, not five different calls to action. Use the same brand, product, and city names on both pages so the relationship is clear. Link back from the landing page under a “Learn more” label for buyers who want the detail before filling out the form.

The landing page keeps the form and proof. The answer page keeps the definitions, figures, and comparison. Let the answer page get quoted; let the landing page get the lead.

Step 4: Can an answer engine read both pages?

I check rendering before rewriting another sentence. Googlebot can run JavaScript, so a client-side page can surface in Google. Other answer crawlers may rely on the initial HTML. In a 41-day test of more than 1,000 pages, the tested bots discovered zero pages linked only through JavaScript. A link that appears in your browser is not enough reason to assume every crawler found it.

Check both URLs in this order:

  1. View the initial HTML. Are the title, H1, offer, price, answer copy, and internal links there? If essential content depends on JavaScript to appear, fix the rendering or generate static HTML before polishing the prose.
  2. Check snippet controls and bot access. A nosnippet tag, a data-nosnippet wrapper around the answer, or a restrictive max-snippet setting can leave you chasing a citation the page is not permitted to earn. That follows from Google’s eligibility rule. I have seen teams spend months editing an answer that sat inside a no-snippet div added years earlier.
  3. Check canonicals and links. Give the answer page and landing page their own canonical URLs rather than pointing one at the other and risking consolidation. Keep the product and city names consistent. Link between the pages in HTML so a crawler does not have to execute JavaScript to find the path.

Skip llms.txt as a citation play. A log study of 137,210 domains found that 97% of published files received zero requests, and Google says Search ignores the file. Fix the readable pages first.

If the bot cannot find the link, read the answer, or quote the passage, another rewrite will not save it.

Step 5: How do you tell whether it worked?

I use a sheet with one row per buyer question and check it weekly in the same browser session. Keep the record plain:

  • The exact query and whether an AI Overview appears.
  • The URL cited, if any: answer page, landing page, or someone else’s page.
  • The passage lifted, so you can see which wording actually answered the question.
  • Search Console clicks to the answer page and analytics data on readers who continue to the landing page and submit the form.

Search Console can help you see page and query traffic. Use analytics to examine the path from answer page to form; do not pretend a rank report proves that path. I put ranking changes in the notes, not the headline, because ranking and citation often split.

Give a change three weeks before judging it. Overview answers change, and a citation that appears Tuesday can disappear Friday without a new edit from you. One check is a snapshot, not a verdict. The useful outcome is a cited answer that sends a buyer toward the form, not a prettier position in a report.

What can software do, and what still takes judgment?

A visibility tracker can show which prompts trigger an Overview and whether your URL appears. A crawler can flag a price trapped in an image or an answer block hidden by a no-snippet wrapper. What the deck calls “AI optimization,” I call spellcheck for structure. Someone still has to decide which question matters, write the plain answer, keep its figures current, publish the page, and repair the template so the next one ships clean.

That execution gap is why I run this work inside groas for clients now. Its models watch prompts daily, rewrite and ship passages, keep internal links and schema aligned with visible copy, and log changes with the reason. A named strategist owns the call on which questions belong on which page. A standalone tool hands you a list. groas does the work around the clock for a flat monthly fee rather than a percentage of spend. Automation is useful when it closes the loop, not when it produces another dashboard to review.

Who should skip the answer page?

Skip it if the landing page already earns its keep from paid traffic and buyers do not ask Google a question before acting. Emergency plumbing, branded searches, repeat orders: those are not all journeys through an Overview. Fix an unreadable offer or broken internal link if you find one, but do not build a second page to satisfy a format nobody in your market uses.

Skip it too if you cannot keep the facts current. A stale price table can do more harm than no citation because the old number can keep circulating after the offer changes. I tell local clients with one crew and a full book to fix rendering, put their useful facts in text, and stop there. Build the second page when the questions keep coming and the landing page can take the leads.

Paper cutout of an answer page with a table linking across a paper bridge to a landing page with a form

What should you do today?

Pick the landing page closest to money. List its top five buyer questions from sales calls, then put its price range and inclusions in plain text without displacing the offer or form. Choose one recurring question that triggers an Overview and publish an answer page for it: direct answer in the first 60 words, useful detail after it, one clear link to the landing page.

Check the citation and the path to the form in three weeks. If that loop works, repeat it. If you want the execution running daily without adding headcount, that is what groas does for clients: passages shipped, links kept clean, and each change logged with its reason.