Does your ad match the landing page? Check it with ChatGPT or Claude

What do Google and Meta say about ad and landing page match?

Google Ads. Quality Score is a 1–10 diagnostic metric for a keyword. It has three parts: expected CTR (the likelihood that the ad will be clicked), ad relevance (how well the ad matches the intent of the search) and landing page experience, which Google defines as how relevant and useful the page is to people who click the ad. An important detail: Google writes that this score is not used directly in the ad auction and should not be optimized as a key metric; it is only a sign of where you can improve (checked October 22, 2026). So this article does not promise either that a fixed page will raise the score or make clicks cheaper.

Google's ad policies are stricter than the metrics. The destination requirements prohibit pages that do not work, pages that return an HTTP error to Google's crawler, and a mismatch between the display domain and the final URL. The misrepresentation policy says that an ad must not promise offers that do not exist, hide the full price or show promotions that are not relevant to the destination, and mismatches between the ad and the landing page count as misleading ad design (checked October 22, 2026).

Meta (Facebook, Instagram). Meta's Advertising Standards say it plainly: the products and services promoted in an ad must match those promoted on the landing page, and all ad elements must relate to the product or service offered. Meta also prohibits deceptive or exaggerated claims about the benefits of a product or service (checked October 22, 2026).

In other words, this check is a simple way to see in time that an ad promises what the page will not deliver.

What to give the AI tool

1. The ad copy as text. For Google Ads: headlines and descriptions (for a responsive search ad, RSA, all the headlines, because Google mixes them), the display path, and assets (e.g. sitelinks, promotion asset). For Meta: primary text, headline, description, the button (CTA) text and whatever is written on the image or in the video. Type out the text from the visual, because that is often where "-20%" or "from €29" hides.
2. The final URL, simply as text, so the AI knows which page is meant and whether it matches the display domain.
3. The landing page text. Copy all visible text (Ctrl+A, Ctrl+C on the page), or save the page as a PDF (in the browser "Print" → "Save as PDF"), or take a few full-length screenshots. If you want the AI to also assess what is visible without scrolling, add a separate screenshot of the first screen on a phone and on a computer. How AI reads PDFs and where it makes mistakes is covered separately (A long PDF with AI). If the offer terms, delivery times or prices are on other pages (cart, "Delivery", "Terms"), upload those too: the person will see them as well.

What not to upload: customer orders, names, email addresses, ad account or website admin access, passwords, API keys. The public page text is enough for this task. If the page is not public yet (a draft), upload only its text without internal notes.

Some ChatGPT or Claude plans can open a link themselves, but they will not necessarily see what a person sees: part of the content, prices rendered by scripts, or the cookie banner may not load. It is more reliable to upload the text or PDF yourself.

A prompt for the comparison

Paste your data in place of the square brackets. The prompt works for both Google Ads and Meta ads.

Why ask for a quote: it lets you quickly check that the AI did not make things up. If the table says "yes" but the quote does not match the claim, treat that row as a mismatch and review it yourself.

Example: an online plant shop ad

Say an online houseplant shop in Panevėžys (a city in northern Lithuania) launches a Meta ad. The example is made up to show what the result looks like. The original copy was in Lithuanian; it is shown here in English.

The ad. Text: "Houseplants from €29, delivered within 24 hours anywhere in Lithuania. -20% on your first order until Sunday!" Headline: "A greener home by tomorrow". Button: "Shop now". The final URL is the "Large plants" category page.

Extract from the page. "Large houseplants. Monstera: €39. Ficus: €45. Strelitzia: €59. … Delivery to parcel lockers anywhere in Lithuania in 1–3 working days. … Subscribe to the newsletter and get a 15% discount code." The page says nothing about a Sunday deadline.

The AI table (shortened) would look roughly like this:

In addition, the AI would note that the headline "by tomorrow" contradicts the page's 1–3 working days. Whether the discount and delivery terms are visible on the first screen cannot be determined from a text extract; the AI should say so and ask for a screenshot.

Still read the table yourself. The AI may miss text that was only in the image, or fail to see that "from €29" could mean smaller plants in another category. It sees only what you uploaded.

What to fix: the ad or the page?

The AI will suggest both options, but the decision is yours. A simple rule: fix the side that does not match reality.

1. If the ad promises more than you can deliver (24-hour delivery when it is really 1–3 working days), change the ad. "Fixing" the page here would just be another promise you cannot keep.
2. If the offer is real but not visible on the page (the promotion is valid but described only in the terms), change the page: show the price, discount and deadline at the top, next to the button.
3. If you are sending people to the wrong page (the ad is about a specific product, but the URL leads to the home page or a broader category), change the final URL, not the copy.
4. If one ad leads to a page with several different offers, consider a separate landing page for that campaign. It is more work, so it is only worth it when the campaign is not short-lived.

How to write new ad copy when you decide to change the ad is covered in the articles on Google Ads copy (Google Ads copy with AI) and Meta ad copy (Facebook and Instagram ad copy with AI). Run the revised copy through the same comparison prompt again.

What to check yourself: the AI cannot see this

Comparing text is only half the job. The AI will not check these things from text or a PDF, so check them by hand before launch.

1. On a phone. Open the ad link on your phone (ideally by clicking the preview from the ad platform). Is what the ad promised visible on the first screen? Is the button hidden further down?
2. The cookie banner. Does the consent bar cover the button or the price? Can it be closed with one tap?
3. Loading. Does the page open without errors and in a reasonable time, especially on mobile data? You can use Google PageSpeed Insights to assess speed.
4. The form or cart. Submit a test enquiry or go as far as the payment step. Does the email arrive, and does the enquiry show up where it should?
5. Tracking. Does the final URL have the necessary UTM tags (URL add-ons such as ?utm_source=facebook that Google Analytics uses to separate traffic), and do they survive a redirect? Does the conversion tag or Meta Pixel fire when the form is submitted? You can check this with the Google Tag Assistant tool and in the test events section of Meta Events Manager.
6. Domain and redirects. Does the domain shown in the ad match where the person actually lands? Google names this directly as a possible reason for disapproval.

This list does not promise better results; it only removes reasons why a good ad might not work.

Common mistakes and limits

Checking only the headline. A Google responsive ad shows different headline combinations, so compare all of them, not just the first. For Meta, do not forget the text on the image and in the video.

Forgotten assets and old promotions. A promotion asset or sitelink with last year's "Black Friday" discount is a classic mismatch. Include them in the comparison.

The page changes, the ad does not. When prices or delivery terms change on the website, the ads stay the same. It is worth repeating the check after every change to prices or terms.

Jumping to conclusions. If a campaign performs badly, a mismatch is only one possible reason. Before changing everything, look at the date range, the spend, whether conversions are measured at all and whether there is enough data. An ad with few clicks proves nothing yet. How to assess this is covered in the article on Google Ads report analysis (Google Ads report analysis with AI).

AI is not a lawyer. AI may notice that "free delivery" has an unstated condition, but it will not assess whether your ad meets consumer protection requirements. In sensitive areas (health, finance), consult a specialist about your claims.

If you have many such checks (dozens of campaigns, frequently changing prices), ad copy and page content can also be compared automatically, through the Google Ads or Meta API and an AI model, for example with an n8n workflow. That is a separate technical integration, worth considering only when manual checking takes too much time.

Sources

Google Ads Help: About Quality Score: the three parts of Quality Score, and that it is a diagnostic tool not used directly in the auction.

Google Ads Help: Landing page experience: definition and where to see the rating.

Google Ads policy: Destination requirements: non-working pages, HTTP errors, domain and URL mismatch.

Google Ads policy: Misrepresentation: offers that do not exist, unclear pricing, irrelevant promotions, mismatches between ad and page.

Meta Advertising Standards: products and services in the ad matching the landing page.

Meta: Unacceptable business practices: deceptive or exaggerated claims about product benefits.