How to Displace a Competitor in AI Answers

To displace a competitor in AI answers, pick one rival and one quarter, then work backward from the answers. Find the grounding searches where they rank in ChatGPT, Gemini, Perplexity, and Google AI Overviews and your client doesn't, open the competitor pages holding those spots, and check which third-party sources keep feeding their name into answers. Split the result into two lists; pages you can write and publishers you have to pitch. A realistic quarter is 8 to 12 target searches, 4 to 6 pages written or rewritten, and 3 to 5 publishers pitched. Track that fixed set in a project so you can show movement against a baseline, and expect most of the change between week four and week ten.

Every agency has this conversation. A client sees a competitor named in ChatGPT and asks why they aren't. The true first answer is that nobody knows yet, and the useful second answer is that it's findable. Somewhere behind that answer is a set of web searches the assistant ran, a handful of pages it read, and a few sources it trusted. Your competitor is in those pages. Your client isn't. That's the whole problem, and it's a much smaller problem than "we're losing at AI."

The trap is treating this like a brand-level fight. Share of voice is a fine scoreboard and a terrible work plan. You can't act on a number that says a rival holds 38% and your client holds 31%. You can act on a list of nine searches where that rival is on page one and your client is nowhere.

This guide works with any AI visibility setup; the callouts show how each step runs in Spyglasses.

Step	What you're doing	What tells you it's working
1. Pick the target	Choosing one rival and one angle	The team can name who and why
2. Find the queries	Listing searches they win and you don't	A fixed target list, 8 to 12 long
3. Read their pages	Studying what makes those pages retrievable	You can explain each win in a sentence
4. Trace the sources	Separating owned wins from earned ones	Two lists, not one vague backlog
5. Build and pitch	Writing pages and pitching publishers	Work shipped against the target list
6. Run the project	Measuring the fixed set monthly	Ranks move on queries you named first

Pick one competitor and one reason

Start by narrowing, because the instinct is to go wide and the result is always the same. A client hands you five competitors, you track all five, and the work never focuses enough to move any of them.

Read the competitor breakdown first and look for the brand sitting one rung above your client. That's usually the right target. The category leader is ahead because of years of coverage, reviews, and links you can't replicate in twelve weeks. The brand just above you is often ahead because of five or six specific pages, and pages are beatable.

Then write down the reason. This part gets skipped and it's the part that makes the rest of the quarter coherent. There are only a few ways a competitor beats you in AI answers, and each one leads to different work:

How they're winning	What it looks like	What the work becomes
They own the searches	They rank for the queries AI runs, you don't	Content and page-level SEO
Third parties carry them	Reviews and roundups name them, not you	Publisher outreach
They win comparisons	Both get named, they get recommended	Positioning and proof content
They're the default answer	AI names them when nobody was asked	Long game, category content

Most clients are losing in exactly one of these ways, and they usually assume it's the last one when it's really the first. Check before you plan.

One more thing worth doing in week one. Compare how often each brand gets named against how often it gets recommended when both appear. A competitor who shows up everywhere but loses head-to-head comparisons is a different problem than one who wins every comparison but barely appears. The first is a visibility gap. The second is a positioning gap, and no amount of new content fixes it.

Find the exact queries where they win

This is the step that turns a complaint into a work plan.

AI assistants don't answer purchase questions from memory. They run web searches, read what comes back, and build an answer from it. Those searches are the real battlefield, and they look nothing like the question the user typed. Somebody asks for the best option for a small team, and behind that the assistant runs four or five searches with different phrasings, comparisons, and qualifiers.

So drop out of the brand view and into those searches. For each one, you want three things: where your client ranks, where the target competitor ranks, and which competitor URL is holding the spot. What comes back sorts itself into three piles.

The first pile is searches where your client already ranks well. Leave them alone. They're working.

The second pile is searches where the competitor ranks and your client sits somewhere past the first ten results. These are your targets. Something exists on your client's site that's close but not close enough, and closing that gap is the cheapest win available to you.

The third pile is searches where your client doesn't appear at all. Some of these are real gaps worth filling. Others are searches your client shouldn't chase, because the intent belongs to a different kind of buyer or a product they don't sell. Be willing to cut half of this pile.

Aim for eight to twelve target searches. Fewer than eight and one bad month makes the whole report look like noise. More than twelve and you can't ship enough work to move any of them. If you're new to reading these searches at all, winning a target AI grounding search walks through the mechanics one query at a time.

Read the pages holding those ranks

Now open them. Actually open them, one at a time, with the target search in mind.

This takes about an hour for a dozen pages and it's the least glamorous hour of the quarter. It's also where the plan stops being generic. You're not asking whether their page is good. You're asking why a retrieval system picked this page for this search.

A few things show up over and over. The page answers the exact question in the first paragraph instead of building to it over eight hundred words. It uses the words the search uses, including the qualifiers, rather than the client's internal product vocabulary. It has a table or a list that can be lifted whole into an answer. It's dated and looks maintained. It loads as HTML without needing JavaScript to render the part that matters.

Write one sentence per page about why it wins. If you can't write that sentence, you don't understand the win yet and you'll write a replacement page that loses for the same reason.

While you're in there, check the boring technical stuff on your client's competing page too. A page that's blocked, slow, or rendered entirely in the browser can be the best answer on the internet and still never get read. That's a fast fix and it's worth ruling out before you commission new writing. An AI readiness audit catches most of it in one pass.

Trace the sources that keep them there

Pages on the competitor's own site are only half the story, and usually the smaller half.

Look at what gets cited in the answers where the competitor is named. You'll see their domain, sure. You'll also see review sites, industry roundups, forum threads, comparison articles, and trade publications. Those third-party sources often carry more weight than anything either brand publishes, because assistants lean on sources that look independent when the question is which option to choose.

Sort what you find into two buckets, because they lead to completely different work and different budgets.

Owned wins are the ones your client can create. A missing page, a thin page, a page that exists but can't be read, a documentation section that answers a question your marketing site avoids. You control the timeline on these.

Earned wins are the ones your client has to be given. A roundup that lists eight tools and not theirs. A review site where the competitor has forty reviews and your client has three. An analyst piece. You don't control these, but they're often the highest-leverage items on the list, and one of them can be worth more than six new blog posts.

Rank the earned list by influence rather than by brand recognition. A trade publication nobody in the client's boardroom has heard of can be cited in answers weekly, while a famous outlet blocks AI crawlers entirely and never shows up. That gap between prestige and actual influence is the thing worth checking before you spend a month chasing a placement.

Build the content and the pitch list

You now have more work identified than one quarter can hold. Cut it down before anyone starts writing.

For the content list, four to six pages is realistic for most retainers. Prioritize by how close your client already is. A page ranking eleventh for a target search is worth more attention than a blank page, because a rewrite that gets it into the top five is a week of work while a new page is a month plus the wait for indexing. Rewrites also come with a track record the search engine already trusts.

Write each page against the sentence you wrote in step three. If the competitor wins because they answer the question in the first paragraph and your client buries it under a brand story, the fix is structural, not more words. Most displacement content is shorter than what it replaces.

For the pitch list, three to five publishers is plenty. Lead with the ones already citing the competitor for your target searches, because they've shown they cover this category and they're missing a source. That's a much easier pitch than convincing a publication to care about the space at all. If your client belongs in a roundup and isn't there, say exactly that and give the writer what they'd need to add them.

Two practical notes for agency teams. First, budget the outreach separately from the content, because the two get done by different people and the pitch list dies quietly when it's a footnote on a content brief. Second, tell the client which pile each item is in. Owned work has a date. Earned work has a probability, and being clear about that up front saves an awkward month three.

Run it as a project and prove the shift

The last step is the one that protects the work, and it has to happen before anything ships.

Record the starting position on your target searches now. Where your client ranks, where the competitor ranks, how often each brand gets named in answers to the tracked prompts. If you skip this and measure later, you'll spend the quarterly review arguing about what the baseline was.

Then set the reporting up around that fixed set of queries rather than around the overall trend. This matters more than it sounds. Overall share of voice moves for reasons that have nothing to do with your work, including competitors launching things, seasonal shifts in what people ask, and the assistants themselves changing how they retrieve. A fixed target list is defensible. A general trend line is a coin flip you have to explain either way.

Read it monthly, not weekly. AI answers vary between runs even when nothing has changed, so a weekly view mostly measures noise and makes cautious clients nervous. Monthly, on the same queries, is the shortest window that says something real.

Expect the shape of the results to be lumpy. Two of your eleven queries will move a lot, five will move a little, and four won't move at all. That's a normal quarter, and it's fine, because the two that moved tell you where to spend the next one. The pattern usually points somewhere specific, like every win coming from rewrites rather than new pages, or every win coming from one publisher placement.

Fold the target-query table into your regular client reporting so it shows up the same way every month. The rhythm for that is covered in building monthly AI visibility client reports, and if you haven't set a starting point for this client at all yet, do that first with baselining a brand's AI visibility.

Take one query at a time

Displacement sounds like a fight and works like a list. One competitor, one quarter, a dozen searches, four pages, four publishers, and a monthly check against numbers you wrote down on day one.

The clients who get impatient with this are the ones who were promised something bigger. Nobody takes a category from a stronger competitor in three months. What you can do is win two or three of the searches that buyers actually run, and then come back next quarter and take two or three more. That compounds, and it survives contact with a client who wants to know exactly what the retainer bought.

Start with the reason. If you can't name how the competitor is beating your client in one sentence, the rest of the plan is a guess with a budget attached.