Skip to content

ResourcesAEO Handbook

From AI visibility gaps to a content plan

Tracking shows where AI answers leave you out. It does not say what to write. Five steps get from one to the other, shown on our own plan.

Jack F, Co-founder

Content strategyAEOAI visibilitycontent gapsGEO

Part of the Answyn AEO Handbook.

Now readingWhat a gap-led plan is

To turn AI visibility gaps into a content plan, sort each gap by what would fix it, update pages the engines already read before writing new ones, pick each page's format from what gets cited for its question, go deep on two or three topics, and brief every page with what it adds. Run on Answyn's 25 tracked questions on 4 November 2026, it found evidence for 22 pieces and set aside 10 gaps no page can fix.

What is a gap-led content plan?

A gap-led content plan is a list of pages chosen from the questions where AI answers leave your brand out, made after each gap has been sorted by what would actually fix it. Tracking shows where you are missing. The sorting decides what kind of work each gap needs, and it is the step most plans skip, which is how a list of gaps turns into a list of articles.

We ran the method on Answyn's own 25 tracked questions on 4 November 2026, using answers collected in the UK from ChatGPT, Google AI Overviews and Claude between 6 October and 4 November. The engines named Answyn in 2 of 1,575 answers. The evidence supported 22 pieces: 17 new pages and 5 updates to pages we already had. A further 10 gaps went on a separate list of things a page cannot fix, among them third-party lists, wrong facts in AI answers, a pricing and access decision, a product coverage gap and a technical check. One of the 25 questions had no honest page answer at all.

That plan is the worked example in each step below.

Step 1: sort every gap by what would fix it

Read the answers for each question you lose, and the pages they cite, then put the gap in one of five groups. Only the first group belongs in a content plan. The rest go on a separate list with an owner, because no page will close them.

Five kinds of fix, with the examples from Answyn's plan of 4 November 2026
What would fix itHow it shows up in the answersIn Answyn's plan
A page of your own, new or updatedRivals are named, and the cited pages are the kind you could publish: product pages, guides, comparisons22 pieces: 17 new pages and 5 updates
A third-party sourceThe cited pages are roundups, videos, posts or forum threads on other sitesOne roundup was cited in 99 ChatGPT runs that left Answyn out. AI Overviews leaned on YouTube, LinkedIn and Reddit
A wrong factThe answer names you and says something untrueChatGPT repeated an offer Answyn no longer makes in 6 answers, logged in Answyn's Brand Facts
A technical problemOne engine never cites you while another reads your pagesAI Overviews cited answyn.com 0 times in 750 answers while ChatGPT read Answyn pages. Cause not yet known; the first check is indexing and snippet eligibility
A product or commercial decisionThe question asks for something you do not sell, or the answers rank on something you do not publishSmall-team answers ranked tools on entry price, which Answyn does not publish. One question asks for engines Answyn's public plans do not collect

The question with no honest page answer sits in the last row. It asks for one dashboard covering ChatGPT, Perplexity and Claude. Answyn's public plans do not collect Perplexity or Claude, so there is no page we could write that wins it without claiming coverage we do not sell. It went on the second list with the reason written next to it.

The second list is still work, and each item on it needs an owner: outreach to the lists that are cited, corrections at the source of wrong facts, technical checks, and the business decisions only someone senior can make.

Step 2: update the pages engines already read

Before writing anything new, look for pages that are read but not named: the engine retrieved one of your pages while researching the answer, then left your brand out of it. Those pages are already within the engine's reach, which makes them the cheapest place to start. Spotting them is one of the checks for why ChatGPT leaves a company out.

Our own case: for the question about which AEO platform a small brand team should choose, ChatGPT read our list of the best AEO platforms for UK businesses in 17 of 30 runs and never named Answyn in those answers. Across five questions it retrieved an Answyn page in 49 runs and left Answyn out of every one. So the first two items in our plan are updates to the two pages it reads most: that list, and our handbook page on the customer journey in AI search.

Read the answers that left you out for the criteria they ranked on, then make sure the page states your position on each one in a passage that makes sense on its own. For the small-team question, the answers ranked tools on entry price, prompt limits and the engines covered at entry level. Our list gave rivals' published prices and "Enquire" for Answyn. An update can state the access route plainly. It cannot publish a price the business has decided not to publish, which is why that gap also sits on the second list.

Finding these pages takes retrieval data, which a citation count alone does not show. For tracked ChatGPT questions, Answyn records the pages ChatGPT retrieved as well as the ones it cited. Its Pages screen also has a "Crawled, never cited" filter, which lists pages an AI crawler fetched that no engine cited, once LLM Analytics is installed on your site.

Step 3: pick the page type from what gets cited

For each question that does need a page, look at what each engine cites when it answers. The cited pages show which kind of page that engine reaches for, and the engines differ.

In Answyn's category, as our study of 1,575 AI answers found, ChatGPT leaned on vendors' own product and pricing pages, on roundups and on official documentation from Google, OpenAI and Microsoft, and on studies with numbers for research-style questions. Google's AI Overviews leaned on YouTube videos and Shorts, LinkedIn posts and articles, Reddit threads and Facebook posts. Every one of youtube.com's 380 citations in the window came from AI Overviews.

That splits the plan by engine. A written page can compete in ChatGPT's answers, because ChatGPT cites the kind of page a company can publish. For AI Overviews in our category a page alone is unlikely to move the answer, so our plan lists short videos, LinkedIn articles and disclosed replies in the threads already cited as off-site work.

This page went through the same step. For the questions we track on this subject, the most-cited domain was developers.google.com, with 94 citations across the engines, almost all of them in ChatGPT, and its cited pages included Google's guidance on creating helpful content. ChatGPT also cited OpenAI's help centre 30 times, and the engines cited how-to posts from Semrush and HubSpot on finding and fixing AI visibility gaps. AI Overviews cited YouTube 68 times, LinkedIn 50 and Reddit 18. The cited posts describe a method in general. This page adds the method run on a real plan, with its numbers.

Step 4: go deep on two or three topics

Group the pieces into topics, then cover two or three of them properly before starting the rest. Answyn's worked model of how much content moves AI visibility puts five to eight pieces into each topic it covers, and argues that one page per gap is the lowest-probability way to spend the same effort. Google's helpful-content guidance makes a related warning when it asks whether you are "producing lots of content on many different topics" in the hope that some of it performs well (Google Search Central).

There is a mechanical reason too. Google says AI Overviews and AI Mode may use query fan-out, issuing multiple related searches across subtopics and data sources to develop a response (Google Search Central). A topic covered by several linked pages gives those searches more to find than one page does.

Questions and pieces do not map one to one. In our plan one study serves four questions, and one page answers two questions about switching platforms. The 22 pieces sit in five topics. After the two updates and the study, the plan works through two topics in depth, then waits eight to twelve weeks before starting the next two.

Step 5: brief, publish, then measure

Write a brief for each piece before anyone drafts it. The brief's main job is to name the information the page adds that the pages already cited lack. Google's helpful-content guidance asks the same thing in two questions: whether the content provides "original information, reporting, research, or analysis", and whether it provides "substantial value when compared to other pages in search results". A page that rearranges what the cited pages already say gives an engine no reason to prefer it.

For Answyn the added information is usually first-hand: our own tracking figures, as on this page, or vendor facts checked on the vendor's own page and dated. Every fact in the brief carries its source, and anything we cannot verify is marked as missing rather than written. Answyn's briefs are built from the collected answers in the same way, and a draft follows only from a reviewed brief. For what other platforms offer, see which AEO platforms turn gaps into briefs or drafts. Access to Answyn is by explicit invitation, so enquire if you want the method run on your own questions.

  1. Add the page's questions to tracking before it goes live, so you have a baseline.
  2. Publish, then give it long enough before judging: about four weeks for a single page and eight to twelve weeks for a topic, as Answyn's published guidance says.
  3. Judge it per engine: whether the page is retrieved, whether it is cited, and whether the answers name you.

Then rerun step 1 on the new answers, because a gap that closes on one engine can stay open on another.

A note on the plan and the data

The worked example is Answyn's own content plan, made on 4 November 2026 from Answyn's tracking of its own category: 25 tracked questions in 8 topics, UK market, 6 October to 4 November 2026, 1,575 answers from ChatGPT (750), Google AI Overviews (750) and Claude (75). It is one brand in one category, so treat the method as the transferable part and the counts as ours.

We sell AI visibility tracking and briefs, so we have an interest in this method. We have tried to show where it says a page is not the answer, including for us, because that is the part a content plan is most tempted to leave out. Google's pages were read on 4 November 2026.

Questions people ask

How many pages do we need to close our AI visibility gaps?
There is no universal number, because it depends on how many questions you track and how often you win a question once it is covered. Answyn's published worked model puts a move from 6% to 20% visibility at roughly 25 to 35 pieces on its central assumptions. For our own 25 tracked questions, the evidence supported 22 pieces, and 10 gaps needed something other than a page.
Should we update existing pages or write new ones?
Update first wherever an engine already reads one of your pages and leaves you out of the answer, because that page is already being retrieved. Write new pages where nothing of yours is retrieved for the question. In Answyn's own plan, the updates to the two pages ChatGPT reads most come before any new page.
What if the pages AI answers cite are on other sites?
Then a page of your own is unlikely to be the fix. If the cited pages are roundups, videos or forum threads, the work is outreach: ask to be added to the list, or take part, with disclosure, where the conversation already happens. In Answyn's category, Google's AI Overviews leaned on YouTube videos, LinkedIn posts and Reddit threads, which no page on our own site replaces.
Do we need special content for AI answers?
Not for Google's, by Google's own account: it says there are no additional requirements to appear in AI Overviews or AI Mode, and no special optimisations necessary. Its guidance on helpful content asks instead whether a page provides original information or analysis, and substantial value compared with other pages in search results. We grade the other common AEO claims in how brands earn citations in AI answers.
How long before we know whether a new page worked?
Answyn's published guidance is about four weeks for a single page and eight to twelve weeks for a topic. Add the page's questions to tracking before it goes live, so there is a baseline to judge it against, and read the result for each engine separately.
Can a tool write the pages for us?
Some tools generate articles at volume. Answyn drafts a page only from a brief you have reviewed, flags claims it cannot back and does not publish to your site. We explain why in why Answyn drafts from a brief, not at volume.

Sources

Run the numbers on your own brand.

Answyn puts the questions your buyers actually ask to every major engine, daily, and records who gets named and who gets cited.