ChatGPT Search crossed 60 million weekly users in April 2026, swapped its Bing-only retrieval pipeline for a hybrid index that pulls from multiple sources, and quietly rewrote the citation ranker. For brands that built a 2025 SEO plan around Google and Perplexity only, ChatGPT just became the third surface that decides whether your name shows up inside an answer.
I am Emily, and I cover the AI search beat for BGR Review. We log ChatGPT Search citations across 3,200 queries every week alongside Google AI Overviews and Perplexity. This is what shifted in the last 60 days, who is winning citation share, and the seven moves to make this week if you want to be the brand ChatGPT picks rather than the brand it summarises around.
What OpenAI shipped in 2026
ChatGPT Search entered 2026 as a feature inside ChatGPT and is exiting Q1 as a serious search product. The user numbers tell most of the story: 60 million weekly users in April, up from 28 million in November 2025, with the biggest spike following the late-March rollout that made search-on-by-default for free users.
The retrieval pipeline changed too. Until February, ChatGPT Search ran almost entirely on Bing's web index. The current build is a hybrid: Bing remains the largest source but OpenAI now mixes in its own freshly crawled pages, partner publisher feeds, and a re-ranking layer that scores candidates on answer fit, source trust and recency. The practical effect is that pages that rank well on Google but not on Bing can now still be cited if the ranker likes the content shape.
OpenAI's chief product officer Kevin Weil confirmed the hybrid retrieval shift in a March keynote. He framed it as moving search from a "Bing wrapper" to "the answer layer of the open web". In April, OpenAI added publisher partner badges to citations from Axel Springer, Le Monde, the Financial Times and the Associated Press, with more deals in negotiation.
ChatGPT Search hit 60 million weekly users in April 2026, up from 28 million in November 2025. The late-March default-on rollout drove most of the growth.
How the new citation ranker behaves
We logged the citation list for 3,200 distinct queries across reputation, local services, finance, health and consumer categories every week through April. The shift after the February ranker swap is clear in the numbers.
Average citations per answer dropped from 5.8 to 3.7. ChatGPT is being more decisive about who it pulls into an answer, and the long-tail sources from the previous Bing-only build are mostly gone. Publisher partners get cited 2.1x more often than equivalent non-partner publishers in the same vertical, which is the steepest preference gap of any answer engine we track.
Recency is now a hard filter for time-sensitive queries. Pages updated in the last 90 days appear in 68 percent of cited slots, up from 44 percent before the swap. ChatGPT also rewards a specific content shape: a clear question-style heading followed by a tight 40 to 80 word answer, then deeper context. That structure gets cited noticeably more often than the same information in long unstructured prose.
The publisher fight that is shaping citations
OpenAI's relationship with publishers is the messiest in AI search. The New York Times lawsuit is still active in 2026, several mid-tier publishers blocked GPTBot in 2024 and 2025, and a small group reversed those blocks after the partner deals went live. The result is a citation pool with visible holes: some authoritative sources never appear because their robots.txt still excludes OpenAI crawlers.
The practical implication for SEO and visibility teams is that ChatGPT often cites a smaller, more partner-weighted set of sources than Google or Perplexity for the same query. If your category is dominated by GPTBot-blocked publishers, the citation slots open up for everyone else, and original primary content from named-author sites can land inside answers faster than it would on Google.
We saw this pattern most clearly in news, finance and tech queries. In categories where the partner publishers do not have deep coverage (local services, niche reputation queries, vertical SaaS), citation share is unusually evenly distributed.
Publisher partners get cited 2.1x more often than equivalent non-partner publishers. Robots-blocked publishers leave citation slots open for everyone else in their category.
Average citations per ChatGPT Search answer dropped from 5.8 to 3.7 after the February hybrid retrieval swap. The ranker is consolidating around a smaller, more trusted set of sources per topic.
Who is winning citation share in May
Across our panel for the first three weeks of May, four groups gained meaningfully. Partner publishers lead, with roughly 2x citation share for their core coverage areas. Named-author expert sites and practitioner blogs are next, particularly in legal, medical, financial and technical categories where ChatGPT defers heavily to recognised expertise.
Government and institutional sources (.gov, .edu, professional bodies) gained citation share for any query touching policy, compliance, or regulated topics. Original-data publishers that ship primary research are gaining across the board: ChatGPT's ranker rewards a unique number with a citation more reliably than it rewards a well-written summary of someone else's number.
The losers are the same as on Google and Perplexity: thin affiliate round-ups, AI-generated content farms, and aggregator sites that summarise other people's reporting. ChatGPT downweights those aggressively, and once a domain pattern is flagged as low-trust, it appears in answers far less often even on its strongest queries.
Your 7-day ChatGPT Search visibility plan
If you want your brand or your publisher pages cited inside ChatGPT Search answers in the next quarter, here is the seven-day plan we run with clients.
Day 1: confirm GPTBot is allowed in your robots.txt. If GPTBot or OAI-SearchBot are disallowed, ChatGPT will not crawl your pages and you will not appear in the citation pool. Then list the 30 to 50 questions a real customer would ask in your category and run each in ChatGPT Search to log who currently gets cited.
Day 2: rewrite the lead block on each priority page. Use a question-style H1 or H2 that matches the user query, follow it with a 40 to 80 word direct answer, then go deeper. ChatGPT's ranker rewards this exact shape.
Day 3: add structured data. Article schema with author and datePublished for everything, FAQPage where appropriate, and Organization with a rich aboutPage. Validate every page in Google's Rich Results Test.
Day 4: refresh dates on the top 20 pages. Update statistics, prices, policy details and the dateModified field. The 90-day freshness window is real and tight.
Day 5: audit your entity surface. Search your brand name in ChatGPT Search. If the answer is wrong, thin or negative, the fix list overlaps with Google and Perplexity: Wikipedia, About page, schema, authoritative third-party profiles, named expert pages.
Day 6 to 7: ship one piece of original primary work per top topic cluster. Original numbers, original quotes from named experts, original screenshots or datasets. ChatGPT cites primary sources at a higher rate than any answer engine we track.
What to watch through summer 2026
Three signals matter for the next 90 days. OpenAI is testing a paid "ChatGPT Pro Search" tier that surfaces deeper answers with more citations and live-data refresh. If that tier ships broadly, citation slots inside Pro Search become some of the most valuable visibility real estate in AI.
Watch the publisher deal list. Each partner badge tilts citation share further toward partners, and the GPTBot block reversals that follow these deals usually arrive within 30 days. If you publish in a category where a major partner just signed, expect your share to compress.
Finally, watch the API. ChatGPT Search's retrieval pipeline is now exposed to developers through the Responses API, which means third-party products are starting to embed it the same way the Sonar API embedded Perplexity. Visibility inside ChatGPT Search increasingly means visibility across a wider answer surface.
