Most content gap advice tells you to run a tool, get a list of questions your competitors answer and you do not, and go write those pages.
Follow that advice and a good share of the pages you write will never get cited. Not because they are bad. Because the gap was never about content.
This is the step where you find out which buyer questions an AI engine is handing to a competitor, and, more importantly, why. The why decides whether the fix is a new page or something that has nothing to do with writing at all.
What a competitor gap in AI search really is
A competitor gap in AI search is any buyer question where an AI engine cites someone else instead of you. There are three kinds, an open gap, a content gap, and an authority gap, and each one needs a different fix. Most teams misdiagnose which one they have and write pages that never get cited. Diagnose first, then act.
The mistake baked into every tool-driven gap report is that it treats all three the same. It shows you a prompt where a competitor appears and you do not, labels it a gap, and implies the fix is content. Sometimes that is right. Often it is not.
Here are the three.
Open gap. You ask the engine a real buyer question and the answer is weak. No strong source, vague hedging, nobody clearly cited. You are absent, but so is everyone good. This is the easiest citation you will ever earn. Write the definitive answer and the engine has nothing better to reach for.
Content gap. A competitor is cited, and you have no page on the topic, or a thin one. This is the classic gap, and content is genuinely the fix. You need a page that answers the question more specifically and with better evidence than theirs.
Authority gap. A competitor is cited, and you already have a solid page on exactly that topic. This is the one that breaks people. The page is not the problem. The engine knows your page exists and is choosing the competitor anyway, because they have something you do not: third-party presence, evidence, a comparison, a Reddit thread, a review-site footprint. Writing another page does nothing. You have to earn the signal the engine is rewarding.
If you cannot tell these apart, you cannot fix any of them efficiently.
How to find the gaps by hand
You do not need a citation-tracking subscription for this. You need the buyer query map you already built and an hour with the engines.
Start with the revenue-ranked questions from your buyer query map. Comparison and alternatives queries first, because those sit closest to a decision. Then work the process for each question.
-
Run the question through the engines your buyers use. ChatGPT, Perplexity, Google AI Overviews, Microsoft Copilot. Run each one three times, not once. These systems are non-deterministic, and a single run is a coin flip. You are looking for what gets cited consistently, not what showed up once.
-
Record who is cited. Write down every source the engine leans on, by name. Your competitors, yes, but also the review sites, forums, and publications. The non-competitor sources matter more than people expect.
-
Answer two questions about yourself. Do you have a page that addresses this? And does the engine cite it? Those two yes-or-no answers are the entire diagnosis.
The two answers map cleanly onto the three gap types:
- No good source cited, you are absent: open gap.
- Competitor cited, you have no real page: content gap.
- Competitor cited, you have a real page that is not getting picked: authority gap.
The diagnostic template
One row per buyer question. This is the version I fill in for a client.
| Buyer question | Engines checked | Who is cited | Do you have a page? | Are you cited? | Gap type | Fix |
|---|---|---|---|---|---|---|
| “[you] vs [competitor]” | GPT, Perplexity, AIO | Competitor blog, G2 | Yes | No | Authority | Earn third-party + add evidence |
| “best X for hourly agencies” | GPT, Perplexity | Nobody specific | No | No | Open | Write the definitive answer |
The “Gap type” and “Fix” columns are the point. A list of missing prompts tells you where you are losing. This tells you why, which is the only thing that tells you what to do next.
Read the citation, not just the ranking
When a competitor wins a question, the reason is usually sitting right there in the answer the engine gave. Read it like evidence.
- Did the engine quote a specific number or claim? “In a study of 40 B2B clients, attributed pipeline rose 34 percent” beats “improves your pipeline” every time. The cited source had a fact. Yours had an adjective.
- Did it pull from a comparison or a table? Engines love structured, side-by-side answers because they are easy to lift. If the cited page has a comparison table and yours has prose, that is your answer.
- Did it cite a forum or review site instead of a vendor? That is the tell for an authority gap. Perplexity in particular leans hard on community sources. If three competitors are discussed in a Reddit thread and you are not, no amount of on-site content closes that.
- Was the cited page recent? Freshness is a real factor. A page last touched two years ago loses to a competitor’s updated one, even if yours was better when it shipped.
The win condition is visible. You just have to look at the answer instead of the rank.
The fix depends on the gap
This is why the diagnosis matters. Same symptom, three different prescriptions.
- Open gap: write the most complete, specific answer in the category. Low effort, fast citation, because you are not outwriting anyone, you are filling a vacuum.
- Content gap: write a page that beats the cited one on the dimension the engine rewarded. More specificity, a real number, a comparison table, a direct answer up top. Match their topic, exceed their evidence.
- Authority gap: stop writing pages. The engine already has your page and passed it over. Go earn what the cited source has. Get into the review sites your category uses, get mentioned in the communities and newsletters your buyers read, add the evidence and comparisons the engine is quoting from competitors. This is off-site work, and it is the gap people waste the most time misfiling as a content problem.
What did not work
A few things I tried that cost time.
Treating a gap tool’s output as a content to-do list. An early version of this process was just exporting a “prompts you are missing” report and writing a page for each. Several of those were authority gaps. We wrote good pages. They did not get cited, because the engine already had pages on those topics from us and was choosing competitors for reasons that lived off our site. The list told us where we lost. It never told us why, so half the work was aimed wrong.
Checking only one engine. I ran the gap analysis on ChatGPT alone and built a plan from it. Perplexity, on the same questions, cited an almost entirely different set of sources, heavier on community and discussion. The gap is not one gap. It differs by engine, and a plan built on one engine is half blind.
Chasing every gap instead of the ranked few. The first maps flagged dozens of gaps. Trying to close all of them spread the effort thin and nothing moved. Closing the three or four gaps closest to revenue, fully, moved the needle. Depth on the questions that sit near a sale beats coverage of every question a buyer might ever ask.
What the gap analysis is for
In our method this is the I in CITE, investigate the gaps competitors are filling. Done properly it does more than tell you what to write. It tells you whether writing is even the right move, which is the part the tools quietly skip because their entire pitch is that you have a content problem they can help you produce your way out of.
You will not. Some of your gaps are content. Some are authority. Knowing which is which is the difference between a quarter of pages that get cited and a quarter of pages that sit there.
Once you know which gaps are real, the next move is choosing which to chase first. That is the Target step: how to decide what to actually publish for AI search.
Run the map. Diagnose each gap. Then spend your effort where the engine has actually left a door open.
If you want this done against your real competitors and your real category, it is the core of a paid audit, and the full method is on the methodology page.