LinkedIn
The pricing page has 84 FAQ answers, 400 lines of JSON-LD, and a badge awarding itself 97/100 for AI readiness. The sentence an AI Overview quoted was there before any of those additions.
Last month a founder sent me a screenshot of a competitor inside an AI Overview and wrote, “Make our pricing page do this.” I know that inbox. Say you are spending $20k a month and the answer you get is a 47-point checklist instead of a paragraph a machine can quote. So I built the page the checklist asked for. Every tip, applied exactly as written.
What follows is the page, with my notes in the margins: the hero line no human would say out loud, the FAQ block, the schema that describes the schema. If you came looking for software that will get your landing page into Google’s AI Overviews, please enjoy the product demonstration.
The brief: “Make our pricing page do this”
The request was simple. The LinkedIn answer was not. Every guide selling a route into AI Overviews seemed to recite the same list: put the direct answer in the first two or three sentences, add FAQ and Article and Author schema, show E-E-A-T, keep the page updated.
I used to tell clients keyword-stuffed ad copy would raise relevance. I was wrong, and I watched Quality Score punish it for months. This felt like the same script with entity pasted over keyword. The vendor promised its software would handle the rest. Below is the finished artifact. Read it the way Google’s crawler would: tired, literal, and in a hurry.
Hero: “As a Leading Trusted Authoritative Provider of Solutions”
Welcome to Our Best Top-Rated Trusted Authoritative Leading Provider Pricing Solutions Page for Businesses, Enterprises, Companies, Firms and Organizations Seeking Affordable Best-in-Class Cost-Effective Pricing Plans and Packages. [Margin note: entity density excellent; readability deceased.]
As a leading trusted authoritative provider of trusted authoritative leading solutions, we provide solution-based solutions that leverage synergies for synergy. What the deck calls “synergy,” I call three adjectives doing the job of one fact. The hero still does not state the price.
Vendor note in the margin: Google understands synonyms, so you do not need to list every variation. That line comes from the people who run the engine, summarized here by a practitioner who admits he used to recommend the hacks. I left the variations in anyway. The checklist said entity density. The engine said it already knows what a synonym is.
The 84-question FAQ, including “What is a question?”
Questions 1 through 84, each with its own H2 and 40-word answer.
- Question 12: What is pricing? Pricing is the amount a business pays for plans and packages. Our affordable cost-effective pricing plans start at a price described on this pricing page. Source: this pricing page.
- Question 13: What is a question? A question is a sentence that asks something. This FAQ answers questions.
- Question 47: Do you offer FAQ schema? Yes. [Margin note: FAQ rich results were restricted in August 2023, then stopped appearing entirely on May 7, 2026. FAQPage markup still validates but produces nothing visible, as this account explains.]
I added the markup anyway because three checklists said pages with FAQ average 12% more citations.
The vendor callout beneath the block reads: Each H2 opens with a self-contained 40-word passage, no forward references, because passages are evaluated in isolation. That part is not a joke. It is structuring advice that prescribes a 30-to-60-word passage with an answer, one data point, and a source. So I did it 84 times. The page now answers everything except what the buyer asked: what will this cost me, and why should I believe it?
Schema on the schema
Below the FAQ sits 400 lines of JSON-LD no visitor will ever see. Organization, WebPage, Article, Author, FAQPage, HowTo, Speakable, Product, Offer, BreadcrumbList, and one proud ItemList that lists the FAQ items for the benefit of a parser.
Google says not to overfocus on structured data because no special schema is needed for AI features. A controlled test of 1,885 pages found much the same: adding JSON-LD moved AI Mode by 2.4% and ChatGPT by 2.2%, both indistinguishable from zero, and AI Overviews by negative 4.6% in the published teardown.
My margin note on the page reads: schema can help a crawler parse the price; it cannot persuade an answer engine to quote it. The distinction would fit nicely in the JSON-LD if we had a field for it.
The /llms.txt love letter
At /llms.txt lives a polite Markdown file addressed to robots that never knock. Dear Claude, Dear Gemini, Dear Perplexity, please find enclosed our pricing in clean Markdown, our brand voice, and a list of approved adjectives.
Googlebot does not use it, it is not a ranking signal, and vendors selling it as a ticket to number one are selling snake oil. The most common failure is the file pointing at URLs that 404, which mine did for a week before I noticed. I kept it because the checklist said future-proof. The engine kept ignoring it because the engine does not read it. A file the crawler ignores is not a substitute for a page it can read.
“Last updated: today” (updated hourly by script)
The byline says Last updated: October 19, 2026. It said the same thing yesterday. A cron job touches the date every hour to signal freshness, while the price table underneath still lists 2024 onboarding fees.
Every freshness guide says keep pages updated, and they are right when something changes. This script changes nothing except the timestamp. Margin note: if the price did not change, the date is decoration. Decoration does not get cited.
The AI Readiness Score badge, self-awarded
Top right, sticky on scroll: AI Readiness: 97/100. Certified Crawlable. Zero errors. The grader that issued it checked for a missing FAQ block and a missing CSP header, then docked three points because my cookie banner was not fashionable enough. It never asked an engine a single question.
Even vendors selling these scores admit the quiet part in the footnotes: high readiness does not mean ChatGPT will cite you; it is not a measure of actual citations. I framed the badge anyway. It converts well on LinkedIn.
The vendor’s results slide
Slide 14 of the deck that sold this page reads: Position 1 pages have a 53% chance of AI Overview inclusion versus 36.9% at position 10. Pages with FAQ average 12% more citations. 44.2% of citations come from the first 30% of the page. The numbers are real, pulled from an 863,000-keyword crawl where top-10 share collapsed from 76% to 38% in eight months.
The conclusion stapled to them is not. Rank helps because retrieval starts with ranking, yet citations also arrive from outside the top 10 through fan-out queries. The slide treats correlation as a work order: add more answers, add more schema, move up three spots. So we did. The page got longer, slower, and less quotable in direct proportion to its score.
The slide has no space for the buyer’s question. I checked. A higher checklist score did not put the price in the answer.
The line the engine actually quoted
I ran the money query through an AI Overview after all of this shipped. It ignored the hero adjectives, skipped questions 1 through 84, and quoted one sentence sitting above the fold that predated every tip. The mechanism is unglamorous: AI Overviews use core ranking and retrieval, and a page must be indexed and eligible for a text snippet before it is eligible for an AI feature. Retrieval found the passage that answered the question in plain language. Everything else was weight.
One answer that existed before any of this
Here is the passage the engine lifted: Plans start at $490 per month for up to 10,000 contacts, onboarding takes 11 days on average, and canceling takes two clicks from settings. Prices exclude ad spend.
It gives a buyer a price, a timeline, a way out, and a catch without sending them to four other sections. It also sits near the top. The average AI Overview runs about 267 words with seven links, and 44.2% of its citations come from the first 30% of the source page. The passage did not need a badge to become useful. Short, early, and self-contained beat long, buried, and cross-referenced here.
Monday’s revision log: remove the optimization
I stripped the page back to what a buyer and a retriever both needed. I deleted 83 of the 84 questions, kept the price sentence at the top, and put the load time back under two seconds. For the longer version of that cleanup, read Stop Writing ‘AI Overview Content.’ Fix the Page Already at Position 8. Its instruction is less marketable than the badge: fix the page that already ranks instead of writing a second page for the robots.
The revision log has three entries. Unlike the FAQ, each one corresponds to something that changed on the page.
Revision 1: put the answer before the equipment
I kept the price, timeline, and catch together in the first paragraph. The short version read: Plans start at $490 per month for up to 10,000 contacts, onboarding takes 11 days, cancel in two clicks. No preamble. No forward reference to a table below. The fuller passage still says prices exclude ad spend, because a buyer should not have to discover that part elsewhere.
Retrieval can lift a passage that stands alone without editing it. And with 44.2% of citations coming from the first 30% of the page, an early answer has a better shot than one buried after the awards ceremony. If you need the buyer phrasing worth answering, read How to Get Cited in Google AI Overviews: Mine the Questions Behind the Query. Margin note: one quotable paragraph beats 84 searchable ones.
Revision 2: let the crawler read the offer
I removed what kept that paragraph from being read. The JSON-LD describing a real offer stayed; the rest went. So did the hourly freshness script and the hero adjectives. What remained was clean HTML, one H1, a price table a parser could read without executing six scripts, and a load time under two seconds on a cold fetch.
This is about eligibility, not elegance. A page must be indexed and eligible for a text snippet before it is eligible for AI features. A crawler that cannot render the price cannot cite the price, however confidently the sticky badge reports Certified Crawlable. Margin note: make the offer readable before making the page look optimized.
Revision 3: check the answer, not the score
I stopped checking the 97 and started checking what got quoted. I ran the three money queries in AI Mode, copied the citations, and compared the quoted line with my page line by line. Position 1 helps, with a 53% inclusion rate against 36.9% at position 10, but rank alone never explained this page’s miss.
The vendor report still looked handsome. If yours does too, compare it with Your Landing Page Scored 94/100 for AI Overviews. The Price Is Still Invisible.. The report measures how completely you followed its instructions. The cited answer tells you whether the engine found something it could use. Those are different jobs. Check the sentence it quotes.
I archived the 97-point version and kept the plain answer. Same URL, same price, a tenth of the markup. It got cited the next week while the longer version never did. The final vendor note is therefore ready for publication: For maximum AI Overview readiness, place one clear answer near the top, then resist the urge to bury it under AI Overview readiness.
I do this work inside groas now, where execution runs around the clock and a human still owns the outcome, and the rule is the same one I learned watching Quality Scores tank in 2016: write the sentence worth quoting first. Everything else is decoration.