Quick answer: a landing page for ChatGPT Ads must continue the conversation the user is coming from: immediately deliver on the ad’s promise, show conditions and evidence, and ask for a clear action. It must also be accessible to OAI-AdsBot, comply with OpenAI’s policies and be tracked with Pixel or Conversions API, with consent managed appropriately. The homepage is rarely the right choice.
On ChatGPT, users reach an ad after explaining a problem, their constraints and often their budget. When they click, they expect the page to pick up that line of reasoning, not start from scratch. This guide explains what OpenAI requires for landing pages, how to structure them and how to track them without losing data. The guidance has been verified as of 5 October 2026.
Why the landing page matters more in a conversational channel
Users arrive with an already detailed question
In a search engine, the query is short; in a conversation, users may already have described who the product is for, when they need it and what they have ruled out. This has two consequences:
- the page must quickly confirm that it is relevant to a specific need, not a generic category;
- information users have already received in the response should not be repeated: they need specific details, conditions and a clear next step.
A generic page forces users to find what they need on their own. In a journey that has been guided up to that point, this creates noticeable friction.
The landing page is a selection signal and is subject to review
For OpenAI, the landing page is not just the click destination. It is one of the signals used to select ads, alongside the conversation context, headline, copy and context hints. It is also examined during review: the policies apply to the entire journey, and an approved ad cannot lead to a page containing excluded content.
Page quality therefore affects both the likelihood of appearing and the outcome after the click.
Technical and policy requirements to check first
Accessibility to OAI-AdsBot
OpenAI uses crawlers to check the safety of pages submitted as ads. The requirement is specific: the landing page must allow access to OAI-AdsBot, which is needed for validation and review. OpenAI recommends also allowing OAI-SearchBot, but identifies OAI-AdsBot as the priority for publishing ads.
The most common blocks to check with the technical team are:
- rules in robots.txt that exclude OpenAI’s crawlers;
- application firewalls and CDNs that treat the crawler as suspicious automated traffic;
- anti-bot systems, human verification checks and CAPTCHAs;
- pages behind authentication;
- overly restrictive rate limits.
OpenAI also publishes its crawlers’ IP ranges for those who need to manage an allowlist of addresses.
Compliance with policies
The advertising policies require the destination to clearly relate to the advertiser and the promoted offer, and to represent them accurately. In practice:
- no exaggerated claims about results, prices or comparisons;
- no content from excluded categories, even in secondary sections reached from the page;
- no elements that imitate the appearance or voice of the ChatGPT interface;
- valid links and a page that is actually accessible.
Recommended structure: the six-block model
The following model is a proposed structure designed for users arriving from a conversation. It is not an OpenAI requirement: it should be adapted to the offer.
Above the fold: an immediate answer
- A headline that reflects the ad’s promise. If the ad refers to “moving a one-bedroom flat in a day”, the page headline must say the same thing, not “Welcome”.
- Essential conditions. Price or price range, timescales, area covered, requirements. These are the first things people comparing options look for.
Evidence, details and objections
- Verifiable evidence. What is included, how the service works, genuine certifications, reviews with sources. No figures that cannot be substantiated.
- Details for those who want to know more. Specifications, use cases, differences between the available options.
- Answers to objections. The questions customers ask before buying: extra costs, timescales, guarantees, what happens if something goes wrong.
Action and form
- A single primary action. Requesting a quote, booking, purchasing or starting a trial. The form should ask only for the information needed for that step. Secondary actions, such as a phone number, should not compete visually with the primary one.
What you remove is equally important. Pop-ups that cover the content as soon as the page opens, navigation menus with dozens of items, autoplay videos and newsletter sign-up requests before users have even read the offer interrupt the journey. People arriving from a conversation have a specific question: any element that does not help answer it should be moved further down or removed.
Messaging: how to continue the conversation
The message should remain consistent throughout the chain: context hint, ad headline, copy, page headline, evidence and call to action. A simple check is to write the chain in five lines and make sure it addresses the same need.
Hypothetical example. An online language school sets up this chain:
- context hint: professionals who need to handle calls with overseas clients within a few months;
- ad headline: English for work calls, individual evening lessons;
- ad copy: a 12-week course with a dedicated teacher and simulated meetings;
- landing page headline: English for work meetings in 12 weeks, with evening lessons too;
- call to action: book an assessment lesson.
If one of the links addresses something else, for example a landing page devoted to language certifications, users perceive a disconnect and the system receives inconsistent signals.
Two approaches help maintain the right tone. First: write as you would respond to someone who has just explained their problem, rather than writing a slogan. Second: do not promise something that ChatGPT’s response may already have ruled out, such as a price you cannot guarantee.
Tracking on the landing page
Pixel, standard events and consent
OpenAI’s Measurement Pixel should be inserted in the head section of the pages where you want to measure conversions. Standard events cover the most common cases:
- leads and registrations: lead_created, registration_completed, appointment_scheduled;
- commerce: items_added, checkout_started, order_created;
- subscriptions and trials: subscription_created, trial_started.
Custom events should only be used when no standard event describes the action.
There is a crucial detail regarding consent: the Pixel treats consent as active by default, unless it is set to false or finds a saved refusal. In Italy, measurement for advertising purposes generally requires user consent: the Pixel must therefore be configured to start without consent and activate only after acceptance. Events blocked before consent are not recovered. If the site uses a Content Security Policy, the domains listed in the documentation must be authorised.
URL parameters and redirects
The Pixel reads an attribution identifier called oppref from the landing URL and stores it in a first-party cookie for 30 days. If a redirect, plugin or CDN rule removes the parameters from the URL, attribution is lost. The same applies to UTM parameters.
Before launch, it is worth running a test: open the landing page with test parameters, follow any redirects and check that the parameters reach the final page intact.
Deduplicating browser and server events
If the same conversion is sent by both the Pixel and the Conversions API, for example a quote request also recorded by the business management system, the same event identifier must be used in both submissions. OpenAI matches duplicate events based on Pixel ID, event name and identifier, counts the first one received and ignores subsequent ones. Without this measure, the same conversion risks being counted twice, distorting cost per result and optimisation.
View-through conversions
When the feature is available for the account, Ads Manager also reports conversions that occur after someone sees the ad without clicking it, with a fixed window of one day. These are a separate metric: the Conversions total, cost per acquisition, bids and optimisation remain based on clicks. Read them as an additional signal of the role played by the landing page and the ad, not as conversions to add to the others.
One landing page for each need, not for each ad
OpenAI recommends linking ads to the most relevant destination, such as product pages, collections or dedicated content, rather than the homepage. It also recommends creating separate ad groups when use cases require different messaging or landing pages.
The resulting rule of thumb is this: one landing page for each distinct need, shared by all the ads in that group. Creating a page for every ad variation multiplies the work without any benefit; using a single page for different needs reduces relevance.
Hypothetical example. A removals company has two groups: flat removals and cellar clearances. Two landing pages are needed because prices, timescales and objections differ. The ad variations within each group, however, all lead to the same page.
Landing page checklist and testing
Before submitting for review
- The page headline delivers on the ad’s promise.
- Prices, timescales or conditions are visible without extensive scrolling.
- The evidence is verifiable and claims are not exaggerated.
- No content from excluded categories on the page or along linked journeys.
- OAI-AdsBot can access the page: robots.txt, firewall, CDN and anti-bot systems checked.
- The page does not require login or hide content behind an interaction.
Before launch
- Pixel installed in the head, consent configured and tested.
- The correct standard event is triggered at the right time.
- URL parameters preserved through redirects and plugins.
- Loading times and readability checked on smartphones too.
After the first few weeks
- Compare the landing page’s conversion rate with that of other channels.
- If clicks are good but conversions are low, work on the page before adjusting bids.
To connect this page to your ad groups, review how to write context hints in ChatGPT Ads.
Frequently asked questions
Can I use the homepage as a landing page?
You can, but OpenAI recommends using the most relevant page, such as a product or service page. A homepage forces users to find the information promised by the ad on their own.
What happens if the firewall blocks OAI-AdsBot?
The page cannot be validated and the ad cannot be published as normal. OpenAI advises correcting blocks in robots.txt, firewalls, CDNs, anti-bot systems, authentication and rate limits.
Do I need a different landing page for each ad?
No. You need a landing page for each distinct need. Ad variations within the same group can lead to the same page.
Does OpenAI’s Pixel comply with privacy consent requirements?
The Pixel provides a consent management function, but the site must configure it. Consent is active by default: in Italy, it must therefore be integrated with the consent management system and checked with a privacy adviser.
How long should the landing page be?
As long as needed to answer the questions that come before the action. A simple offer may fit into a few sections; a service with many variations requires details and answers to objections. What matters is that essential information is immediately visible, not the total length.
Does the landing page affect cost?
The landing page is one of the signals used for ad selection, which takes place through an auction weighted by relevance. A page that is consistent with the ad supports relevance and, after the click, conversion. For a complete analysis of the journey, from the page to brand visibility, see the AI Search Visibility & ChatGPT Ads service.
If you want to check that your landing pages are ready for ChatGPT Ads before investing, talk to Digital Creation about your strategy.
Sources and updates
Updated as of: 5 October 2026
Did you enjoy this article?
