What AI content deals mean for your AEO strategy
When an AI company signs a content deal with a publisher or a platform like Reddit, does getting your brand mentioned there become more valuable? For anyone working in answer engine optimization (AEO), that is the strategic question: should these agreements change where you invest in content, PR and community participation?
The answer depends on what the deal enables and what happens in AI answers afterward. OpenAI’s Reddit partnership provided structured access to current discussions, yet researchers later measured a sharp decline in Reddit citations. Access and visible prominence can move differently, so a partnership announcement alone is insufficient reason to prioritize a source.
Before you change your strategy
- Check the scope: training rights, live access and attribution affect different parts of an AI system.
- Check the evidence: an announced deal does not establish a citation uplift.
- Check your market: use repeated buyer questions to decide which sources deserve your effort.
What does an AI content deal actually change?
Start with the rights involved: an archive for training, a feed of current articles, or permission to show summaries and links. These serve different purposes.
| Agreement covers | What it can enable |
|---|---|
| Model training | Use of content when developing or improving models. |
| Live content access | Retrieval of current articles, discussions or information. |
| Display and attribution | Summaries, excerpts, quotations or links inside a product. |
| Compensation | Payment or other benefits for agreed uses. |
A deal can combine these rights. Check the announcement for the affected product and permitted uses.
From a content agreement to an AI answer
- 01Access
The system can obtain the content.
A license or data feed can change this. - 02Selection
It chooses sources for the question.
Check what appears for your audience. - 03Representation
It describes a brand and may cite a page.
Check the answer, attribution and link.
The major content deals behind AI search
The early agreements focused on archives. In July 2023, the Associated Press and OpenAI agreed to collaborate, including access to selected archive content. AP is a news agency that supplies reporting to other newsrooms and publishes at AP News.
By 2024, the announcements also covered current discussions and attributed journalism. Google and OpenAI gained structured access to Reddit; the Financial Times and News Corp signed agreements with OpenAI. In 2025, the Times’ Amazon agreement showed how training and real-time display rights could sit in the same deal.
| Date and parties | Confirmed scope | What it means for AEO |
|---|---|---|
| July 2023Associated Press + OpenAI | Access to selected archive content and technology collaboration. | Check for live retrieval and display rights before treating an archive license as a new media opportunity. |
| February 2024Google + Reddit | Expanded access to structured, current Reddit information through its Data API. | Check which current discussions appear for questions about your category. |
| April 2024Financial Times + OpenAI | Licensing partnership including attributed summaries, quotations and links to FT journalism. | Investigate the FT pages appearing for your market before prioritizing outreach. |
| May 2024Reddit + OpenAI | Real-time, structured Reddit content access alongside product collaboration. | Track specific threads and source changes, rather than relying on Reddit’s overall prominence. |
| May 2024News Corp + OpenAI | Access to content across participating News Corp publications. | Evaluate the participating titles individually for relevance to your audience. |
| May 2025NYT + Amazon | Real-time summaries and excerpts, plus foundation-model training. | Evaluate the publication separately for each AI product. |
For AEO, this history raises a more useful question than who signed the biggest deal: which of these sources actually contributes to answers in your market? Reddit gives us a way to compare an access announcement with later behavior.
Reddit: access expanded, then citations shifted
The 2024 agreements
Google’s February 2024 announcement described expanded access to Reddit’s Data API. OpenAI’s May announcement described access to real-time, structured Reddit content to help its products understand and surface community information.
Reddit remained prominent well after the agreement: Semrush found it cited in close to 60% of tracked ChatGPT responses in early August 2025, more than a year after the May 2024 announcement. The later declines describe changes in citation behavior, not an end to the partnership.
“OpenAI will access Reddit’s Data API, which provides real-time, structured, and unique content from Reddit.”
Reddit stayed prominent after the deal, before later citation shifts
- OpenAI + Reddit
Structured access to current discussions.
Partnership announced - Reddit widely cited
In close to 60% of Semrush’s tracked ChatGPT responses.
2025 study - A sharp reduction
Around 10% of responses in the same study.
2025 study - Another decline
Promptwatch and Otterly observe new drops.
2026 study
What changed in the citations?
Semrush recorded a sharp shift in 2025. Promptwatch and Otterly observed another in August 2026. Their measurements use different samples and units, so read each comparison on its own.
Semrush · 2025
Reddit appeared in fewer tracked responses
≈60% → ≈10%
Share of tracked ChatGPT responses citing Reddit
- Window
- Early August to mid-September 2025, within a July 14–October 12 study.
- Sample
- More than 230,000 prompts, weekly snapshots over 13 weeks across three AI platforms.
- Finding
- Reddit’s presence in ChatGPT responses fell sharply in the observed period.
- Limit
- Response-level appearances, not share of all citations. Proposed explanations, including a Google search-parameter change, remain unresolved.
Promptwatch · 2026
August 2026: a later drop in Reddit citation share
3.83% → 0.52%
Share of ChatGPT citations attributed to reddit.com
- Window
- July 18–August 7 versus August 14–17, 2026.
- Sample
- Aggregated ChatGPT interface observations; the article does not specify an exact sample count for these windows.
- Finding
- An 86.4% relative decline between the specified comparison windows.
- Limit
- The authors explicitly say a collection issue cannot be ruled out.
Otterly · 2026
A separate sample found fewer daily citations
497 → 132
Average daily Reddit citations across 16 brand reports
- Window
- August 6–13 versus August 14–17, 2026.
- Sample
- 16 US brand reports. Daily Reddit citation counts fell 73.4%; overall ChatGPT citations rose 3.5%.
- Finding
- The decline in this sample was not simply a fall in total ChatGPT citation volume.
- Limit
- Only four closed days in the post-change window. Different prompts, windows and denominator from the other studies.
What could explain the later declines?
2025: access to search results changed. Google removed num=100, a parameter tools used to request more results. If ChatGPT depended on that route, fewer Reddit pages might have reached its answers. Semrush discusses this explanation but questions whether it accounts for the whole decline; its researcher also suggested efforts to reduce over-reliance on particular sites.
2026: ChatGPT may have changed how it searched. Otterly points to reports of more searches targeted at specific websites, alongside gains for official and reference sources in its sample. That could favor known sites over forum threads discovered through broader searches. It is a proposed explanation, not a confirmed account of OpenAI’s retrieval system. Promptwatch also flags possible collection issues.
The AEO lesson is that a source can stay available through a partnership while its share of citations changes. Review which sources are gaining for your questions before moving effort away from Reddit.
Check whether the shift affects your strategy
Suppose an agency invested in Reddit because relevant threads kept appearing for a client’s buyer questions. Before changing that investment, it should:
- Re-run the client’s established buyer questions.
- Check whether Reddit’s presence changed in those answers.
- Compare engines separately, then identify the sources gaining presence.
- Check whether the client’s mentions, description, referrals or qualified enquiries changed.
- Adjust effort when the evidence warrants it, and diversify excessive reliance on one source.
Use the same checks for your own brand. If Reddit still contributes useful answers and qualified interest, keep that evidence in the decision. If other sources are replacing it, investigate those opportunities.
THE NEW YORK TIMES
The New York Times: a dispute with one company, a deal with another
The Times sued OpenAI and Microsoft in December 2023. In May 2025, it licensed content to Amazon for summaries and excerpts, including uses in Alexa, and for training foundation models.
The same publisher can take different paths
A publication’s relationship with Amazon tells you little about its relationship with OpenAI. Evaluate a media opportunity against the AI product your audience uses and the questions where the publication appears.
Also check the cited page itself. Tow’s attribution research found answers linking to syndicated or copied versions of articles. Coverage in an original publication and a link to that publication are not always the same outcome.
What to check before pitching an outlet
- Permitted use: read the agreement or authoritative announcement.
- Observed presence: check repeated, dated answers for relevant questions.
- Business value: examine whether coverage contributes to accurate brand representation and useful outcomes.
Do licensed publishers receive more AI citations?
Otterly and Press Ranger compared citation patterns for licensed and unlicensed publishers. In their June 2026 ChatGPT sample, licensed publishers received more citations per cited page.
Otterly and Press Ranger · June 2026
More citations per cited page in the licensed group
Publisher size, authority and subject matter may help explain the difference. The calculation includes cited pages, rather than all published pages. Deal records also need to be matched to their effective dates: this study’s citation window is June, while its deal evidence extends through July 28.
Use licensing status as one input when researching an outlet. Give more weight to its relevance and its actual appearances for your audience’s questions.
Frequency and accurate attribution are different outcomes
Tow Center’s March 2025 study tested whether eight AI search tools could identify news articles from excerpts. Content partnerships did not guarantee correct attribution; some answers pointed to syndicated or copied versions. That article-identification task differs from typical buyer questions and from a frequency study.
Four outcomes to check
- Is the publication cited?
- Does the cited page mention your brand?
- Does the answer accurately describe your brand?
- Does that visibility produce a useful visit or enquiry?
Apply the evidence to your market
A publisher deal can matter less than a directory listing for a local service, or less than a specialist review for a software buyer. Start with the questions people ask about your category.
| Business type | Sources worth investigating |
|---|---|
| B2B and professional services | Specialist publications, associations, expert sources and relevant professional platforms. |
| Ecommerce brands | Editorial reviews, buying guides, comparisons and customer-experience discussions. |
| Local businesses | Location pages, directories, review platforms and local publications. |
| Publishers and content brands | Original articles, syndicated versions and sources reusing or describing their work. |
Local search shows why context matters. Yext’s location-based research reported 44% of citations from first-party websites and 42% from listings in its sample. It examined 6.8 million citations across ChatGPT, Gemini and Perplexity, four industries and location-specific query types during July–August 2025. Those proportions should not be generalized to every market.
Turn the source map into an AEO plan
Use the same loop as our practical AEO guide: map recurring sources, correct claimable sources, earn credible inclusion, then re-run and compare.
Map recurring sources
Choose a stable set of buying questions about providers, comparisons and selection criteria. Run them repeatedly on the platforms your audience uses. Record the date, location, cited pages, brand mentions and factual accuracy.
Correct what you can manage
Update the owned pages, profiles and listings appearing in the source mix. Fix outdated descriptions, product information, location details and conflicting facts.
Earn credible inclusion
Offer expertise, original evidence and useful resources to publications and communities your audience trusts. Prioritize relevance and observed source inclusion when choosing where to contribute.
Re-run and compare
Repeat the questions after substantive changes. Compare source inclusion, brand representation, referrals and qualified outcomes. Use unchanged questions or pages as a comparison where possible, so platform-wide shifts are easier to spot.
Avoid relying on one source
Review how much observed visibility depends on one publication, platform or engine. Diversification decisions should follow your source map and audience, rather than a universal channel checklist.
How to read the next AI content announcement
Use these six questions to decide whether an announcement changes your priorities.
| Ask | Check |
|---|---|
| What changed? | Separate training rights, live access, display permissions and measured source selection. |
| Where? | Identify the specific AI product, market and experience. |
| What is established? | Distinguish an announcement, a rollout, an observed change and a causal explanation. |
| Does it affect our questions? | Check relevance to your audience and source mix. |
| What can we influence? | Identify a useful change to owned information or earned presence. |
| How will we verify it? | Define a dated baseline, repeated observations and useful outcomes. |
Understand the sources shaping your audience’s questions, improve the information you can influence, and revisit the source map as relationships and systems change.
For the wider process, including technical access, content and measurement, read the complete AEO guide.
Sources
Compiled by Team AI Page Ready from the announcements and studies below. Last reviewed October 27, 2026. The citation figures describe the original study periods, rather than current market share.
We have summarized the published findings, not independently audited the underlying datasets.
- AP and OpenAI collaborationPrimary announcement · July 2023
- Google’s expanded Reddit partnershipPrimary announcement · February 2024
- Financial Times and OpenAI partnershipPrimary announcement · April 2024
- OpenAI and Reddit partnershipPrimary announcement · May 2024
- News Corp and OpenAI agreementPrimary announcement · May 2024
- The Times announces its OpenAI and Microsoft lawsuitPublisher’s reporting · December 2023
- The New York Times and Amazon licensing agreementPrimary announcement · May 2025
- Semrush: The Most-Cited Domains in AICitation study · published November 2025
- Promptwatch: Reddit Citations Are Dropping in ChatGPTCitation study · published August 2026
- Otterly: ChatGPT cut Reddit citations by at least 73%Citation study · published August 2026
- Otterly and Press Ranger: AI News Licensing DealsLicensing and citation study · published August 2026
- Tow Center: AI Search Has a Citation ProblemAttribution study · published March 2025
- Yext: location-based AI citation researchCitation study · published October 2025