Competitor ad research: what rival ads and landing pages tell you
Ad libraries show you what competitors say, not whether it works. A teardown method that ends in a decision about your own landing page and offer.
- Competitor ads show choices, not results. You can read the offer, promise, proof and call to action, and whether the landing page keeps the promise the ad made.
- Spend, conversion rate and cost per lead stay private. No public ad library shows them, so a long-running ad is a hint about what the advertiser keeps paying for, not proof that it works.
- The method is a grid, filled the same way every time. Four public sources, six fields per ad and a dated snapshot each month. AI is useful for summarizing and comparing what you collected.
- The output is a decision about your own page. In the worked example, a better message match moves the lead rate from 2% to 2.5% and cuts cost per lead by 20% at the same cost per click.
A competitor’s ads and landing pages tell you what they offer, what they promise, what proof they show, what they ask the visitor to do, and whether the page keeps the promise the ad made. They do not tell you what the competitor spends, how well the page converts or whether the campaign makes money. So the useful output of competitor ad research is not an ad to copy. It is a decision about your own landing page and offer.
That distinction matters because the half of paid media you fully control is on your website. The auction sets the price of the click. Your landing page sets how many of those clicks become a lead, and that rate decides your cost per lead.
This guide sets out a teardown method you can repeat: which public sources to use, what to record, what you may infer and where to stop. It keeps the part of the original article that still holds, which is using AI to summarize and compare what you collect. The conversion event throughout is a lead, meaning a demo request or an enquiry. The teardown example is an invented category, and no real advertiser is described.
What competitor ads and landing pages reveal
An ad and its landing page are a competitor’s public answer to four questions: what do we sell, why should you care, why should you believe us, and what should you do next. Those answers are choices, and choices are visible. The results of those choices are not.
What the research is really for sits in the middle of the paid funnel. Your cost per lead is your cost per click divided by your landing page conversion rate. You and your competitors bid in the same auctions, so the click costs you all a similar amount. The difference between accounts is made after the click.
The gap between the two rows is message match. When a visitor clicks an ad about one thing and lands on a page about another, part of the click is wasted before the offer is even read. A competitor teardown is a fast way to see how others in your category handle that step, and where they leave it undone.
Google has its own landing page experience rating for the same link. Google Ads Help (read 6 October 2026) lists three components of Quality Score. Ad relevance is “how closely your ad matches the intent behind a user’s search”, and landing page experience is “how relevant and useful your landing page is to people who click your ad”. The same page states that “Quality Score is not an input in the ad auction”, so read it as a diagnostic, not as a price. Which attributes describe a good landing page experience goes through that rating in detail.
Four public sources and what each one shows
You need four sources, and all four are public: the Google Ads Transparency Center, the Meta Ad Library, the LinkedIn Ad Library and the landing pages the ads point to. The first three are run by the platforms themselves, which makes them the only competitor ad data that is observed and not estimated.
- Google Ads Transparency Center. Google’s advertising policies help (read 6 October 2026) calls it “a searchable repository of advertisers and the ads they’ve served on Google platforms like Search, Display, Gmail, and YouTube”. You can search by advertiser or website name and filter by date and targeted location.
- Meta Ad Library. When Meta opened the library to every advertiser, its announcement (28 March 2019) said it was bringing in all active ads any Page is running, not only political ones, to create one place to see them.
- LinkedIn Ad Library. LinkedIn’s transparency page (read 6 October 2026) describes it as “a searchable collection of ads that have run on LinkedIn”.
- The landing pages. Open the destination of each ad in a clean browser, on desktop and on a phone. This is the half of the teardown most teams skip, and it is the half that relates to your own website.
Third-party tools that estimate a competitor’s keywords, budgets or traffic are a different kind of evidence. Their numbers are modeled. This method uses only what you can see with your own eyes, so every line in the grid can be checked by someone else on the team.
If your competitors do not advertise, or win mostly through partners, marketplaces or outbound sales, the libraries will be close to empty and tell you little. The same is true of a new category with one or two advertisers. In those cases the landing page comparison is still worth doing, and the ad part is not.
What to record for every competitor ad
Record six things for each ad: the promise, the offer, the proof, the call to action, the message match between ad and landing page, and how long the ad has run. Use the advertiser’s own words. A grid filled with quotations can be compared. A grid filled with impressions cannot.
Invented category: fleet maintenance software, sold by demo
- Promise. The outcome the ad headline offers, copied word for word.
- Offer. What the visitor gets for converting: a trial, a demo, a quote, a price list, a download.
- Proof. What supports the promise on the page: named customers, measured results, reviews, certifications. Note where it sits. Proof below the fold is not seen by everyone.
- Call to action. The wording of the main button and what happens after the click, including how many form fields stand between the visitor and the lead.
- Message match. Whether the page headline repeats the promise in the ad, and whether the offer in the ad is the offer on the page. Rate it strong, partial or weak, with one sentence of evidence.
- How long it has run. The dates the library shows for the ad where it shows them, plus your own dated snapshots.
The last field needs a routine, because a single visit to a library is one frame of a film. Run the same searches on the same day each month and save what you find. After a few months you have something no single visit gives you: which ads were kept and which were dropped.
Decide which competitors go in the grid before you start. The original version of this article sorted them into three groups, and that still works: direct competitors who sell a similar product to the same buyer, indirect ones who solve the same problem another way, and aspirational ones with a stronger brand and a larger budget. Two or three from each group is enough. Benchmarking only against companies that look like you produces a grid where every column says the same thing.
What you can infer and what you cannot
You can infer what a competitor has decided to say and to whom. You cannot infer whether saying it works. Every field in the grid is an observation. The step from observation to conclusion is where competitor research usually goes wrong, so it helps to write the limit next to each one.
| What you observe | What you can reasonably infer | What you cannot see |
|---|---|---|
| The promise in the ad | The problem and the buyer they have chosen to lead with | Whether that promise wins the click, or their click-through rate |
| The offer on the page | How much commitment they ask for at the first step | How many visitors accept it: their conversion rate |
| The proof they show | Which objections they expect from the buyer | Whether the proof changes behavior on the page |
| Message match between ad and page | How much care goes into the step after the click | What a mismatch costs them, or whether they have noticed |
| How long an ad has run | That they chose to keep paying for it | Spend, cost per lead, lead quality or profit |
| How many ads and variants are live | How actively they test creative | Budget, bids, targeting detail or which variant gets the traffic |
The nearest public substitute for the missing column is a benchmark, and it is a poor one for this purpose. Ruler Analytics (26 May 2026) reports an average conversion rate of 5.4% for paid search and 2.11% for paid social across more than 110 million sessions, counting a qualified lead or sale as the conversion. That tells you what is typical across 13 industries. It tells you nothing about the one advertiser in your grid.
Common mistake: treating a long-running competitor ad as a proven winner and copying it. You do not know its results, and even if it works for them, it was written for their offer, their price and their landing page. Copy it and you compete on their promise with a page built for yours.
There is a second reason to stop at the limit. Statements about a named company’s ads, performance or intent are claims you would have to stand behind. Keep the grid internal, keep it descriptive and keep conclusions about your own page.
Where AI helps: summarizing and comparing
AI is good at the slow part of a teardown: reading thirty ads and ten landing pages and telling you what they have in common. It is the part of the original article that holds up best. A language model will summarize, cluster and compare text faster and more evenly than a person on a Friday afternoon.
It does not add evidence. A model given a competitor’s landing page knows what the page says. It does not know what the competitor spends or how the page converts, and if you ask, it may produce a confident answer anyway. Give it the copy you collected and ask only questions the copy can answer.
Paste the ad text and the page text yourself. A summary is only as complete as what went in, and a page the model could not load is a gap you will not see in the answer.
- Ad copy, word for word
- Landing page copy, in page order
- Your own ad and page, same format
- Promise, offer, proof and call to action per ad
- Themes that repeat across competitors
- Where each page fails to repeat its ad
- Checking each answer against the source
- Anything about spend or results
- The decision
Prompts that stay inside the evidence
These are adapted from the prompts in the original article, with the questions a model cannot answer taken out.
- Extract. “From the ad and landing page copy below, list the core promise, the offer, the proof and the call to action. Quote the exact words for each. If one is missing, say it is missing.”
- Check the match. “Does the landing page headline repeat the promise made in the ad? Is the offer in the ad the same as the offer on the page? Answer strong, partial or weak, and quote the two lines you compared.”
- Compare. “Across these competitors, which promises appear more than once? Which buyer objections does nobody address?”
- Turn it on yourself. “Here is our ad and our landing page in the same format. Where is our message match weaker than the examples above? What do we offer that none of them mention?”
- Track change. “Here are this month’s and last month’s snapshots for the same advertiser. What changed in the promise, the offer or the proof?”
One prompt from the original is worth keeping for the landing page alone: ask what assumptions the page makes about the buyer’s priorities, and whether the copy shows any sign of understanding the buyer’s objections. It forces a reading of the logic behind the page instead of a description of it.
The useful output of competitor ad research is not an ad to copy.
What a better message match is worth
If closing a message match gap lifts your landing page conversion rate from 2% to 2.5%, your cost per lead falls by 20% at the same cost per click. The formula is the one behind the whole PPC cluster: CPA = cost per click / conversion rate. The guide to how conversion rate changes CPA and ROAS works through it in full. Here it is applied to one decision.
The arithmetic, step by step. 1,000 clicks at $6 cost $6,000. At 2%, those clicks produce 20 leads, so each lead costs $6,000 / 20 = $300. At 2.5%, the same clicks produce 25 leads at $6,000 / 25 = $240.
The lift in conversion rate is 25%. The cut in cost per lead is 20%, because CPA moves by 1 / 1.25. To get the same $240 from the media side, the cost per click would have to fall from $6 to $4.80 with no loss of volume or intent.
The lead here is a demo request or an enquiry. If sales accepts a smaller share of the new leads than of the old ones, the gain shrinks, so check cost per qualified lead as well. Why paid lead generation brings volume but not qualified buyers covers that check.
To put real numbers beside the invented ones, the WordStream by LocaliQ 2026 Google Ads benchmarks (updated 19 May 2026) cover 13,474 US search campaigns that ran between 1 April 2025 and 31 March 2026.
The same formula holds in the published averages: $5.42 divided by 8.18% is $66.26, within a dollar of the reported $66.69. WordStream counts the conversions advertisers track in their own accounts, so its rate is not comparable with the Ruler Analytics figure above, which counts a qualified lead or sale. Use the benchmark whose conversion event matches yours, then use your own numbers.
Turn the teardown into a landing page decision
A finished grid should produce a short list of decisions about your own landing page and offer, each with a measure attached. If it produces a list of ads to imitate, the research has answered the wrong question.
The original article gave three examples of this step, and they translate directly into page decisions:
- Competitors hide pricing behind a demo. The decision is whether your landing page shows a price. The ad can then promise it truthfully, and the page must deliver it above the fold.
- Competitors claim a lot about AI and show little. The decision is to replace general claims on your page with a specific workflow or a recorded walkthrough.
- Features are level, and setup is where buyers struggle. The decision is to put the setup time and the first steps on the page, with proof.
In each case the ad changes last. Write the page so it keeps the promise, then write the ad that makes it. How to optimize a landing page without hurting UX covers the build, and the traffic or website diagnostic helps you check whether the page is the weak term in your account in the first place. Before you read any before and after result, make sure the clicks are human: AI agents clicking ads can distort both rates.
Where Pathmonk fits
Pathmonk works on the step after the click. It reads first-party, cookieless behavior during the visit, predicts the buying stage (awareness, consideration or decision) and serves stage-matched microexperiences without changing the page content and without dev work. Lift is measured against a control group, which is how you find out whether a decision from the grid changed your lead rate. On the campaign side, the Pathmonk Ads agent builds campaigns and landing pages from the same first-party behavior data.
For the research itself, Pathmonk’s Competitor Spy page describes a feature that maps your market from your Company DNA and analyzes rivals across SEO, AI search and PPC, with ad creatives drawn from public ad libraries. Where it shows ad spend, the page labels it as an estimate, which is consistent with the limits set out above.
A teardown cannot tell you what your buyers want. It shows what competitors believe they want. If the whole category makes the same promise, the grid will show agreement, not truth. Evidence about your own visitors, from your own traffic, outranks anything in a competitor’s ad.
Key takeaways
- A competitor’s ads and landing pages show the offer, promise, proof, call to action and message match. They never show spend, conversion rate or results.
- Use four public sources: the Google Ads Transparency Center, the Meta Ad Library, the LinkedIn Ad Library and the landing pages themselves.
- Record six fields per ad in the advertiser’s own words, and take a dated snapshot each month to see which ads are kept.
- A long-running ad shows a choice to keep paying for it. It is not evidence of profit and it is not a template.
- AI is useful for summarizing and comparing the copy you collected. It cannot supply numbers that are not in the copy.
- CPA is cost per click divided by conversion rate. In the hypothetical example, moving from 2% to 2.5% cuts cost per lead from $300 to $240.
- The output of the research is a decision about your own landing page and offer, tested on your own traffic.
FAQs about competitor ad research
How can I use AI to reverse-engineer my competitors’ ad strategy?
Collect the ads from the public libraries (Google Ads Transparency Center, Meta Ad Library, LinkedIn Ad Library) and the copy of the landing pages they point to. Then ask a language model to extract the promise, offer, proof and call to action from each, and to group the ads by theme. That shows what competitors have chosen to say. It does not show their spend or results, and a model should not be asked to guess them.
Can you see how much a competitor spends on ads or their conversion rate?
No. The public ad libraries show ads, not commercial ad budgets, cost per lead or conversion rates. Third-party figures for a competitor’s spend are estimates. The closest public reference is an industry benchmark, which describes an average and not the advertiser you are looking at.
What prompts should I use to analyze a competitor’s landing page?
Use prompts the page copy can answer. Ask what assumptions the page makes about the buyer’s priorities, whether the copy shows evidence of understanding buyer objections, and whether the page headline repeats the promise in the ad. Ask for exact quotes with each answer so you can check them against the source.
How do I decide which competitors to analyze?
Sort them into three groups: direct competitors with a similar product for the same buyer, indirect competitors that solve the same problem another way, and aspirational ones with a stronger brand and larger budget. Two or three from each group is enough. Comparing only against companies that look like yours produces little contrast.
Can AI help me monitor changes in competitor strategy over time?
Yes, if you keep dated snapshots. Save the same ad library searches and landing pages each month, then give a model two snapshots for the same advertiser and ask what changed in the promise, the offer or the proof. Over several months the snapshots also show which ads were kept and which were dropped.
How do I write better ad hooks using AI competitor research?
Do not imitate the competitor’s message. Ask a model to summarize the promises that repeat across the category, then write hooks around something your landing page can prove and the others do not mention. Build the page first so it keeps the promise, then write the ad.
What is message match and why does it affect cost per lead?
Message match is how closely the landing page repeats the promise and offer of the ad that sent the click. Cost per lead is cost per click divided by conversion rate, so a page that converts more of the same clicks lowers the cost of each lead. In the hypothetical example in this article, a move from 2% to 2.5% at a $6 click cuts cost per lead from $300 to $240.
- Google Advertising Policies Help, ads transparency (read 6 October 2026) (support.google.com)
- Meta Newsroom, A better way to learn about ads on Facebook (28 March 2019) (about.fb.com)
- LinkedIn, Our Transparency Center (read 6 October 2026) (about.linkedin.com)
- Google Ads Help, about Quality Score (read 6 October 2026) (support.google.com)
- WordStream by LocaliQ, Google Ads benchmarks 2026 (updated 19 May 2026) (wordstream.com)
- Ruler Analytics, conversion rate benchmarks 2026 (26 May 2026) (ruleranalytics.com)