Competitive Analysis for Paid Media: How To Find the Broken Stage

TL;DR

Competitive research can show which stage of your funnel is broken (Be Seen, Be Believed, or Be Chosen) and which channel to fix it in.

• Ask where you're weaker than a competitor, not what they do well
• A Be Seen gap shows up as mentions and impressions competitors earn that you don't
• A Be Believed gap shows up as weaker reviews, case studies, or AI answers than a competitor has
• A Be Chosen gap shows up when their landing page beats yours for the same query
• Score any new channel against the specific gap it needs to close before you pitch it
• Prove upstream spend without perfect attribution using four proxies: direct traffic, attributed branded searches, geo holdouts and incrementality tests, and LLM citation tracking

Most of us use competitive research the same way. We pull a few ad examples, check auction insights, steal a headline idea, and tweak a bid. That's optimization, and it's fine.

It's also just the tip of the iceberg at what competitive research can do for you.

The bigger use is diagnostic. If you look at competitors the right way, they'll tell you which part of your funnel is actually broken and which channel to go fix it in.

Here's the framework I use, how to run it, and what it looked like when we tested it on ChatGPT ads.

B2B Buyers Don’t Want to Talk to Sales, and Mismatched Messaging Is Part of Why

Start with why this matters. Gartner surveyed 632 B2B buyers in 2024, and 61% said they prefer an overall rep-free buying experience.

Put plainly, about 6 in 10 B2B buyers would rather not talk to a salesperson at all.

So why do we think that is? The same survey gives us a clue.

69% of buyers reported inconsistencies between the information on a seller's website and what the seller told them.

Gartner doesn't say that's the cause, but it points at trust.

If what buyers read on your site doesn't match what they hear later, they'd rather skip the conversation and do their own research.

That means a lot of the decision happens before your sales team, and often your paid search campaigns, get a chance to weigh in. If your program only shows up at the bottom of the funnel, when someone searches your brand or clicks a bottom-funnel ad, you may be competing for a spot that you never had a fighting chance at.

Trust is the stage most programs skip, so that's where I want to start.

Three Stages, Three Different Problems: Be Seen, Be Believed, Be Chosen

Here's how I break the journey up. Each stage is its own problem, and most paid strategies only fund the last one.

1. Be Seen. Does the right buyer know you exist before they ever search?

I'm not talking about impressions. I'm talking about a qualified presence in front of the right person before they start comparing vendors. Most teams can tell me their impression volume. Very few can tell me what those impressions did.

1. Be Believed. When they find you, through an ad, a landing page, an organic result, or an AI answer, do they believe you can solve their problem?

This is the squishy one, and most reporting stacks are blind to it. It's still measurable: retargeting conversion rate, whether your ad message matches the landing page and offer, and whether lead quality is improving over time instead of just volume.

1. Be Chosen. Are you turning that visibility and trust into action?

This is the stage everyone already measures: cost per acquisition, cost per sale, return on investment.

These stages work as a chain. Be Seen builds the audience pool, Be Believed qualifies the right buyers, and Be Chosen converts them.

If you skip one, the next stage has to work twice as hard. If "Be Chosen” looks healthy but the other two are flat, you're capturing demand that already existed instead of converting demand you created. That's a more fragile spot than it looks like on a dashboard.

It also explains why CPAs creep up for no obvious reason. The cause is often two stages upstream.

Skip one stage and the next has to work twice as hard

 

Be Seen

Does the right buyer know you exist before they ever search?

Builds the audience pool

 

Be Believed

When they find you, do they believe you can solve their problem?

Qualifies the right buyers

 

Be Chosen

Are you turning that visibility and trust into action?

Converts them

seerinteractive

Ask "Where Am I Weaker?" Instead of "What Are They Doing Well?"

Most competitive research starts with the wrong question: what is my competitor doing well, and how do I copy it?

Try these questions instead: where am I weaker than my competitor? Is it visibility, credibility, or conversion? And in which channel?

That second question does two things the first one doesn't. It tells you which stage is broken, and it points you at the specific channel to fix it in.

How to Spot a Gap at Each Stage

The Be Seen gap: competitors earn mentions and impressions you don't.
Your standard competitive tools (Semrush, AdClarity, SimilarWeb) help here, but don't stop at the ad library screenshot. Look at who shows up in communities, press, and other places your brand tracking doesn't cover. If a competitor has a meaningful presence somewhere you have none, that's your first candidate.

The Be Believed gap: competitors are described more favorably than you.
Look at review presence, ratings, case studies, and press mentions. Then ask ChatGPT, Gemini, Perplexity, and Claude your comparison questions and see who gets described better. If your visibility goes up but your belief signals don't move, people are seeing you and deciding you're not worth their time. That needs a different fix.

The Be Chosen gap: their offer wins at the point of decision.
Put your landing page next to theirs for the same query. If you're losing there, no amount of upstream work will fix it.

What Counts as White Space (and What Doesn't)

White space isn't just any channel where competitors aren't spending. That could mean it doesn't work for your category.

I'm also not chasing the newest, trendiest platform, and I'm not copying what a competitor just launched. I'm looking for a channel where my buyer already is, that I'm not in, and that closes a specific gap I've already diagnosed.

Five Ways to Find White Space

1. Map search and conversation behavior. Find where the conversation about your category already happens (forums, subreddits, community groups) and compare it to where paid media is running. A loud conversation with silent ads is worth investigating. Tools like SparkToro can show where your ideal customer actually spends time, which gets you from impressions to being seen by the right people.
2. Ask your platform reps. They have vertical benchmark data you'll never see in a dashboard: category CPMs, what similar brands are testing, where the platform is pushing new inventory, and peer set reports on media mix. Most teams never ask.
3. Scroll manually. Never discount the power of a manual search. Build profiles that match your target, scroll the feed, and log what you see. Automated tools miss placements and formats that only show up when a human is looking. I love a good screenshot of what someone sees in the SERP or an LLM result.
4. Read what public competitors tell investors. They talk about marketing efficiency and channel mix on the record because they have to. Not many people in paid media read these. Pair them with competitor reviews, sales calls, and Reddit sentiment, which is real people describing real problems.
5. Check AI answer visibility on a schedule. Ask the models your category comparison questions and see who gets mentioned and recommended. Whoever's missing is the white space. This moves fast, so it belongs in your regular routine, not a one-time audit.

Score the Channel Before You Pitch It

Once you've found a candidate, I run it through a weighted scorecard tied to the gap I'm trying to close:

• Audience presence. Is your buyer actually there? I weigh this one heaviest, at 35%.
• Platform mechanics fit. Does the ad format support your funnel? If you don't have CRM integration for LinkedIn lead forms, that format isn't a great use of budget for you.
• Competitive absence. Is it empty because nobody's found it, or because it doesn't work?
• Cost efficiency hypothesis. Do you have a defensible reason to expect efficiency before you have data? Have you budgeted time for testing and learning?
• Internal capability (a blocker flag, not scored). Do you have the creative, tracking, and bandwidth? A channel can score well on the other four and still be the wrong bet this quarter. If you don't have UGC video creative, TikTok isn't realistic for you right now.

Two channels can score the same and solve different problems. A channel with high audience presence and poor mechanics might fix a Be Seen gap and do nothing for Be Chosen. Score it against the problem you're actually solving.

Four Proxies for Proving Upstream Spend Without Perfect Attribution

The channels that fix “Be Seen” and “Be Believed” are the hardest to attribute, which makes them the hardest to fund. Perfect attribution is a myth, and we've written about that before. These proxies hold up in budget conversations:

1. Direct traffic and revenue. Look for movement against your upstream spend over 60 to 90 days, depending on your conversion window.
2. Attributed branded searches. This is a newer Google conversion action that links video exposure to a brand search within seven days. I wrote up how it works. It's not perfect, but it's a leading indicator.
3. GEO holdouts and incrementality tests. Use geography as a control group and test whether your spend is actually incremental or just stealing from somewhere else. Use the data to prove the value.
4. LLM citation tracking. Track prompts to see whether you're cited, mentioned, or recommended, and whether competitors are. Treat it as a Be Believed health metric.

Teams that win this argument aren't waiting for perfect attribution. They build proxies that are directionally right and defensible.

What a Client's ChatGPT Data Showed: Seen and Believed, but Rarely Chosen

Here's how this played out in a channel I'm piloting right now.

For one client in the home-improvement space, we tracked 253 non-branded category prompts across about 15,900 ChatGPT answers from January through July 2026. The pattern was lopsided:

• Be Seen: The brand appeared in at least one answer for 252 of 253 prompts. In individual answers it was closer to 4 out of 5.
• Be Believed: About 85% of mentions were positive, about 12% were mixed, and fewer than 1% were negative.
• Be Chosen: Only around 5% of answers actually recommended the brand. We estimated that with a recommendation-phrase check, so treat it as a ballpark.

ChatGPT names the brand, describes it well, and then doesn't tell the reader to pick it.

A lot of brands measure the first one and report it like the third.

Seen and believed, but rarely chosen

Be Seen answers that mention the brand

80%

Be Believed mentions that are positive

85%

Be Chosen answers that recommend the brand

 

About 5%

Based on 253 tracked prompts (about 15,900 ChatGPT answers, Jan to Jul 2026). Be Seen and Be Chosen are shares of non-branded answers. Be Believed is the share of brand mentions that were positive. The recommendation rate is an estimate.

seerinteractive

Build to the PROMPT, Not the Page

In a normal search campaign, I start with the landing page and the keywords that get people there. In ChatGPT, the unit of competition is a question someone asks, and the answer next to your ad is part of the competition.

That changes the question. If an answer describes you badly, you probably don't want an ad next to it. So ask what job the ad needs to do for that prompt.

Here's how I'm approaching it:

1. Cluster your prompts by intent. Research phase, comparison phase, and so on. It's the same thinking as the n-gram analysis we used to do on search term reports.
2. Map each cluster to a stage and a gap. Then match the ad variation to that gap.
3. Start where the citation gap is loudest. That's your clearest white space.

This isn't hypothetical. We found competitors buying ChatGPT ads on exact prompts where one of our clients was losing, including switch and comparison prompts. Our own R&D lead's early view of ChatGPT ads shows it's early too: only 7.3% of branded-prompt responses carried the brand's own ad.

Five Jobs an Ad Can Do, Depending on How You Show Up

Same channel, five different jobs:

1. You never appear. Buy visibility. Earn a seat at the table, maybe with CPM bidding, especially if you're up against big brands.
2. You show up less than half the time, with little rivalry. Build presence. Buy consistency before you try to win an argument.
3. Rivals are present and you're rarely on top. Tip the decision. Many ChatGPT answers end with something like "if you tell me X or Y, I can narrow this down." That's your opening to answer the question before the user wraps up the conversation.
4. Rivals are present and you're usually on top. Draw the contrast. Show why you beat them so nobody gets swayed the other way.
5. You're always on top with no rival. Protect the spot against a competitor ad, or move that budget to a category where you have a real gap.

If you can't say which of these you're in, you're not ready to set a bid. Go figure out how you show up first.

What Our ChatGPT Ads Pilot Taught Us About Which Ads Earn the Click

We ran a month-long pilot for the same client, and the winners lined up with one idea: does the answer leave the user something left to do?

• Answer already complete, high demand (cost questions). ChatGPT just gives a number, so the cost calculator ad got a low click-through rate. Here the ad is a visibility play. Judge it on submissions or presence, not clicks.
• Answer leaves a job unfinished (choosing a product line, visualizing a design). These did best. The model lists options without choosing, or can't render a design, so the ad finishes the work.
• Intent isn't native to the surface (find a local builder or retailer). These were the worst performers. People don't ask a chatbot to find a contractor down the street. That's a map behavior.

One more thing: organic ChatGPT traffic ran about 2.3x the paid traffic and engaged more. Paid brought in far more first-time visitors. So paid is actually reaching colder audiences at the moments the answer stops short.

Here’s What You Can Do About This

1. Audit which stage is broken. Use competitors as your diagnostic and run the Be Seen, Be Believed, Be Chosen checks. Where are you showing up, and where are you falling short?
2. Pick one white space channel and score it before you pitch it internally.
3. If you test ChatGPT ads, plan around the job the ad needs to do, not the fact that the channel is new. The same philosophy applies to other channels too.