How JEV gets your pages ready for Google and AI search
JEV scores how ready each page is to rank in Google and be quoted by AI assistants like ChatGPT. See the 11 steps, and what it found on a client site.
Google result
The page that used to win the click
now also has to be
AI answer
The passage an assistant can quote
The page that used to win the click now also has to be the passage an assistant can quote.
Buyers still search Google. More of them also ask ChatGPT, Google’s AI Overviews, and other assistants for a short answer and a few sources. The page that used to win the click now also has to be the passage an assistant can quote.
JEV is Atomic Design’s review for that job. It scores how ready each page is to rank in Google and to be quoted by AI assistants like ChatGPT. You see the weak spots, the first fix, and whether the work changed anything.
This is not a count of citations you have already earned, and it is not a promise of a higher ranking. It is a page-by-page readiness review, in a fixed order, with a second read after the work.
Below is how the eleven steps work, in depth, and what JEV found when we ran it on a national B2B services company. The company is not named here.
Why this matters now
A buyer can get an answer without visiting your site. If your offer is buried, your claim is vague, or a fact on the page does not match the rest of the web, an assistant can skip you and quote the page ahead of you instead.
Traditional SEO still matters. A page still needs a clear title, a useful heading, enough substance, and a path a crawler can reach. AI search adds a second test: can a machine lift a direct, current, provable answer without distorting it?
JEV reads the pages you already have and shows both sides of that test: what a searcher can use, what an AI answer can quote, and what to fix first. Eleven steps. The first read finds the gaps. The later steps say what to change. A person publishes. JEV does not.
How JEV works: the 11 steps
Each step below follows the same shape: what happens, why it matters, an example, and what you get.
- 01Read the live site
- 02Score every page
- 03Line up the fixes
- 04Keep or change each element
- 05Match the questions buyers ask
- 06See the chance of being quoted
- 07Sort the searches you already earn
- 08Connect the pages that belong together
- 09Compare the pages ahead of you
- 10Check a draft before it is published
- 11Read the site again
1. Read the live site
What it does. JEV starts from your sitemap and reads the main commercial pages. If the sitemap is thin, it follows links from the homepage. Blog posts, news, legal pages, contact pages, and utility pages stay out of the default set. A large site is read in a bounded pass, and the review says when pages were left for later.
Why it matters. The findings have to come from pages a buyer can actually open, not from a staging copy, a leftover PDF, or a page that only exists in a content calendar.
Example. On a service business, that usually means the homepage, the service pages, location pages, and the key decision pages that support a quote. A privacy policy or a thank-you page is not the first place to look.
What you get. The pages in the review, and a note when the read stopped early.
2. Score every page
What it does. Each page is checked for a useful title and search description, one clear heading, supporting sections, a canonical URL, a noindex tag, internal links, enough substance to evaluate the offer, helpful buyer detail, trust evidence, and an answer that is easy to quote. Image descriptions are checked when the page has images.
Those judgments come from the words on the page. Claims are not verified off the site, and citations you have already earned are not counted. The noindex check reads the tag on the page. It does not read robots.txt or the server headers.
Why it matters. Google and an AI answer both need a page that states the offer and can be quoted. A vague page loses both. A noindex tag, a missing heading, or a title that does not match the offer is a specific problem, not a vague “needs SEO” note.
Example. Two service pages can look finished and still fail in different ways. One may have a weak title and no direct answer in the opening. Another may be well written but quietly marked noindex, so search and AI retrieval never use it, or should not use it.
What you get. A finding on each page, with the fix and why it helps, and a page-by-page view of what passed.
3. Line up the fixes
What it does. Findings stay on the live pages. Fixes that repeat across the site are grouped apart from fixes for a single page, then sorted so the highest impact comes first.
Why it matters. An unsorted audit is a pile. Clearing one template issue, a repeated title pattern, a missing FAQ block, a sitewide schema gap, can move every page before anyone rewrites a single URL.
Example. If fifty location pages share the same thin heading pattern, that is one template fix, not fifty unrelated tasks. A single leftover noindex on a private page is a different kind of item and should not get lost in the pile.
What you get. A list you can work from the top. A person ships the edit and checks the live page while signed out, with a before and after. JEV does not publish the change.
4. Keep or change each element
What it does. The title, the search description, the main heading, the FAQ, and the schema each get a keep or a change. A change comes with a suggested replacement.
Why it matters. These are the lines a search result and an AI answer read first. A duplicated title or a missing FAQ is a specific edit, not a rewrite of the whole site.
Example. A service page might keep a strong H1, change a generic meta description, and add an FAQ that answers “What does it cost?” and “How far in advance should we book?” without inventing a price that has not been confirmed.
What you get. A decision for each element, and replacement text when the decision is change.
5. Match the questions buyers ask
What it does. Buyer questions are matched to the best page you already have. Questions with no page are listed on their own. For a question with no page, JEV can check who an AI answer cites instead.
Why it matters. One service page cannot answer every question. The gap is a missing page, not a guess. If buyers ask “how does implementation work?” or “how should we compare options?,” and that answer only lives in a sales call, an assistant will quote someone else.
Example. A national services site may already cover “what does this service include?” on a guide page, while “how do we measure ROI?” still has no clear owner. The review should say which question already has a home and which one does not.
What you get. Questions that already have a page, questions that do not, and who is cited when your site is not the match.
6. See the chance of being quoted
What it does. Each page gets a read on how likely it is to be quoted, plus the single first fix. That fix might be a direct opening, one sourced fact, an FAQ, a comparison table, or a visible date. The read looks at how directly the page answers, how fully it covers the decision, what it can prove, how current it looks, and whether it is easy to scan.
Why it matters. An AI answer quotes the clearest passage. This is a chance, not a count of citations you have already earned, and it is not a promise that you will be quoted.
Example. A cost guide that opens with a real range, names what is included, and shows when the page was updated is easier to quote than a brochure paragraph that says “affordable, professional service.” The second page may be pleasant to read and still give an assistant nothing safe to repeat.
What you get. The first change that would make the page easier to quote.
7. Sort the searches you already earn
What it does. The queries your site already appears for are marked as a buyer search or not. A buyer search is one a person uses to hire, buy, book, or compare. Jobs, definitions, and logins are set aside.
Why it matters. The pages to fix first are the ones close to a purchase. A definition query and a “hire [service] in [city]” query should not compete for the same attention.
Example. “What is this service?” is useful educational traffic. “Hire a [service] provider in [city]” is a buyer search. Both can be worth ranking for. Only one is close to a quote request, and that is the page you usually want to strengthen first.
What you get. Buyer searches separated from the rest. This step waits until those queries can be read.
8. Connect the pages that belong together
What it does. Each page is compared with the pages closest to it in subject. A link and anchor text are suggested only when the topics honestly overlap.
Why it matters. The page you want to rank should receive links from the pages that share its subject. A guessed link does not help, and a generic “learn more” click does not tell Google or a reader why the two pages belong together.
Example. A city guide should point to the matching service page for that city, and a “how should we compare providers?” guide should point to the service page that can actually fulfill the work. A random leap from a payment policy to a city page is not a useful link.
What you get. Proposed links, and a pass when a link would only be a guess.
9. Compare the pages ahead of you
What it does. A competitor page is scored on four things: whether it answers directly, whether it goes deep enough, whether it offers proof, and whether it looks current. JEV then checks where that page ranks in Google for the search term, and whether ChatGPT cites it.
Why it matters. You can see the page a buyer meets instead of yours, and what that page does that yours does not. Ranking and citation are related, but they are not the same test.
Example. A competitor may rank well and still be weak on proof, or look current and still fail to get cited. Another page may be the one ChatGPT cites because it opens with a direct answer and a fact a machine can lift. The four-part read, Answer, Depth, Proof, and Freshness, makes that visible without guessing.
Whether it answers directly.
Whether it goes deep enough.
Whether it offers proof.
Whether it looks current.
What you get. The four-part read, the Google position for that term, and whether ChatGPT cites the page.
10. Check a draft before it is published
What it does. A draft has to pass twenty checks before a person reviews it. It must name the audience and the offer, open with a direct answer, include a real fact, answer a buyer question, and close with a calm next step. It must not invent a rank, a review score, a guarantee, a medical or legal result, or an insult. Placeholder text fails. A price, a name, an address, or a phone that has not been confirmed is not published.
Why it matters. A published fiction is worse than a missing page. In AI search, a wrong claim can be repeated. The check exists to stop that before a person hits publish.
Example. A new city page that still contains “lorem ipsum,” an invented review score, or an unconfirmed hourly rate should fail the check and stay in draft. The page can go live once a person confirms the facts and the failed checks are cleared.
What you get. A draft cleared for a person, or the checks that failed. Nothing goes live from this step.
11. Read the site again
What it does. After the fixes and any new pages, JEV reads the site again and sets that read next to the first one.
Why it matters. The first read is a baseline. The second read shows what changed in the answers, the proof, and how easy the pages are to quote. Without it, you only have a to-do list, not a result.
Example. If a site already scores 99+ on the first official read and 99+ on the next complete read, the headline is not a dramatic leap. The value is the page list, the remaining actions, and a confirmed second score, not a promise that traffic already moved.
What you get. A before and after. The second read is still not a count of AI citations, and it is not a promise of a higher ranking.
What you receive
A finished JEV review gives you:
- The pages in the review, plus a note when the read stopped early
- A finding for every page: what passed, the fix, and why that fix helps a searcher or an AI answer
- A fix list in impact order, with template fixes separated from one-page fixes
- Keep-or-change decisions for the title, search description, main heading, FAQ, and schema, with replacement text when the decision is change
- Buyer questions matched to a page, questions with no page, and who an AI answer cites for the gaps
- A citation chance and a first fix, not a count of citations you already earned
- Buyer searches separated from the rest, once those queries can be read
- Suggested internal links and anchors only where two pages share a subject
- A competitor read: Answer, Depth, Proof, and Freshness, plus the Google position and whether ChatGPT cites that page
- A draft that is ready for a person, or the twenty-check failures that still need work
- A second read, set next to the first one, so you can see what changed
Who it is for
JEV is built for owners who need the site to bring in leads, teams watching buyers ask AI first, anyone about to publish a service or answer page, and both local and national businesses. A local page still has to show one consistent name, address, and phone. A national site still has to make the offer easy to quote.
What JEV will not do
- Count the citations you have already earned
- Publish a page, a price, or a fact that has not been confirmed
- Promise a ranking or a citation
- Invent a fact that is not on the page
Case study: a national B2B services company
This case study is about a national B2B services company. The company is not named, and the industry is not identified, so a competitor cannot recognize the business. We ran JEV on the live site and recorded what it found, including what was still open after the complete official read. It is not a traffic-results story. The complete official run was recent enough that later clicks cannot be treated as a JEV outcome.
What JEV scored
A recent official read covered about 400 pages and scored 99+, with a grade of A. An earlier official baseline on the same site was also 99+, also grade A.
Two things are worth saying plainly.
First, a score of 99+ is already strong. This was not a broken site waiting for a rescue. The site already had a large commercial inventory, clear service pages, and a substantial set of buyer guides.
Second, a high score is not the same thing as “nothing left to do.” After the complete official read, JEV still listed about 20 open fixes across roughly 15 pages. One of those was a High Technical noindex finding on a private page. A noindex tag on a private page is often the correct choice. The point of the finding is that the review caught it and put it on the list for a person to confirm, instead of leaving a private URL buried in a site of about 400 pages.
The page count moved slightly across the official reads. That is why this draft does not treat the second score as a perfectly matched before-and-after of the same URL set. It is a second official score, on a complete read of about 400 pages, with the remaining actions still visible.
- ~400 pagesA recent official read of the live site
- 99+ scoreGrade A on both official reads
- ~20 fixesOpen across roughly 15 pages
What this shows about JEV
JEV is useful on a weak site because it tells you what to fix first. It is also useful on a strong site, because it still finds the leftover issues a person can miss: a private page that should stay out of the index, a short list of URLs that still have open actions, and a second score you can compare with the first one.
That is the honest version of this case. The site scored in the A range on the first official read, completed a rescore of about 400 pages at 99+, and still had a short, specific action list. No ranking or citation gain is claimed from those reads.
Traffic context, not a JEV result
The company already receives organic search traffic, and some visits are attributed to AI assistants. Organic visits roughly doubled from June to October.
Those figures describe the site’s existing traffic. They are not presented as a result of the recent JEV runs, and they are not a forecast. If we later measure a clean after period, this section should be updated with dates and sources, not with a retroactive claim.
Questions worth asking first
Is JEV a count of how often AI cites us?
No. It scores how ready a page is to be quoted. It can also check whether ChatGPT cites a specific page. It does not tally the citations you have already earned.
Will it change our website?
It recommends the edit and checks the draft. A person publishes. A price or a fact that has not been confirmed stays unpublished.
Which pages are included?
The main commercial pages from your sitemap. Blog posts, news, legal pages, contact pages, and utility pages stay out of the default set. If the sitemap is thin, JEV follows links from the homepage. A large site is read in a bounded pass, and the review says when it stopped early.
What if we cannot share our search queries?
The page scores, the element decisions, the citation chance, and the link suggestions still run. Sorting buyer searches waits until those queries can be read.
Does a strong citation chance mean we will be quoted?
It means the page is easier to quote. It is not a promise of a ranking or a citation.
Does this check the whole technical setup?
It checks the page itself: the title, headings, canonical, and whether the page tells search engines not to index it. It does not read robots.txt or the server headers.
Why is the case study anonymous?
Because a named company could be recognized by a competitor. The example is a national B2B services company. A recent official read covered about 400 pages, scored 99+, and still left about 20 open fixes across roughly 15 pages. That is a better proof of how JEV works than a made-up before-and-after.
Did JEV already increase this company’s traffic?
Not in any way we can honestly measure yet. The complete official run was recent. Existing organic and AI-assistant traffic is included above as context only.
See what JEV finds on your site
If you want the same read on your pages, tell Atomic Design which site to start with. You get the pages, the fixes, and a second read after the work. Nothing is published until a person reviews it.
Sources & attribution
- Atomic Design. This article describes JEV, Atomic Design’s review of how ready each page is to rank in Google and to be quoted by AI assistants. A person publishes. JEV does not.
- Case study. This case study is about a national B2B services company. The company is not named, and the industry is not identified, so a competitor cannot recognize the business.
- Traffic context. Organic visits roughly doubled from June to October. Those figures describe the site’s existing traffic. They are not presented as a result of the recent JEV runs, and they are not a forecast.
Let's build the
next thing.
Tell us where you're going. We'll show you the system that gets you there, and the team that's been doing it since 1996.
Got it, thank you!
We will be in touch shortly.