Lureon Data Brief · Monthly Series AI Search Monthly · July 2026

July 2026 in AI Search: ChatGPT took 92.4% of assistant referrals while 40 to 60% of citations rotated out from under everyone

Four things changed in July: the traffic picture consolidated onto one platform, Google finally gave marketers AI visibility data, ChatGPT opened a self-serve ad auction, and the evidence that AI answers destroy valuable clicks got harder to argue with. This is Lureon's monthly brief on what happened, what it means, and what to do before the next cycle.

Covering: 1 to 31 July 2026 · Published 2 August 2026 · Next edition: early September

Free 30-minute call · No retainer required to talk

The assistant market consolidated onto one platform

Previsible published its third AI Traffic Study in early July, analyzing 6.77 million LLM-driven sessions across 166 GA4 properties over 19 months. It is the longest-running dataset in the category, and it tells a consolidation story.

Three takeaways. Claude grew 64x and overtook Perplexity in March, concentrated among developers, technical buyers, and professional services, which is exactly the B2B profile our Claude platform brief covers. Perplexity and Copilot show that last year's consensus bets can reverse fast, so single-platform strategies are fragile by construction. And the study's own caveat is the strategic headline: AI discovery inside Google's own surfaces almost certainly moves more traffic than every standalone assistant combined, which keeps Gemini and AI Overviews the priority surface even though ChatGPT wins the referral count.

One operational detail worth acting on this month: roughly 28.8% of ChatGPT's referrals land on a site's internal search results page rather than the page that answers the question. The model trusts the domain but cannot always identify the right URL. Internal site search has quietly become an acquisition surface, and almost nobody has optimized it.

Google gave marketers real AI visibility data, with gaps

Search Console's generative AI performance reports expanded to more sites through early July, after launching in June. This is the first native view of visibility inside Google's AI surfaces.

Also in July: Google switched on AI Overviews and AI Mode in France on 22 July, the last major market without either surface, so French impression curves start from zero right now. And the European Commission issued binding Digital Markets Act decisions on 16 July requiring Google to open Android features to rival assistants and share anonymized search data with competing engines and chatbots, which over time widens the set of surfaces a brand needs to be visible in.

ChatGPT became a two-layer channel

On 22 July, OpenAI opened a self-serve Ads Manager, letting advertisers build campaigns targeted on conversational context rather than keywords. The answer now has an organic layer and a paid layer, the same split Google Search went through, compressed into a single launch.

What this does not change: the recommendation itself is still earned. Paid units sit beside the answer, labeled as sponsored, while the brands named inside the answer get there through authority signals. If anything, a visible ad layer raises the value of organic citation, because buyers discount what they can tell was bought. The deeper shift is agentic commerce: when an assistant makes a purchase recommendation on a user's behalf, it draws on structured product and entity data, not campaign creative. Clean feeds and consistent entity data are becoming an acquisition asset.

The click-loss argument got harder to dismiss

An updated randomized field experiment published in July measured a 39.8% reduction in organic clicks on searches where AI Overviews appeared. The methodology matters more than the number.

Google's standing defense has been that AI answers mainly absorb low-value or accidental clicks. The updated study tested exactly that: the extra visits generated when Overviews were removed showed no statistically significant difference in return-to-search behavior, visits under ten seconds, or time on the destination site. The lost clicks were normal clicks. The first organic result nearly doubled its clicks when the Overview was removed, and losses concentrated on informational queries, where Overviews appear most. Our standing analysis of this shift lives in AI Is Eating Search.

Downstream, the economics are visible elsewhere: publisher ad inventory fell by roughly a third to two fifths year over year in the second quarter as zero-click experiences reduced referrals. For B2B brands the lesson is not to mourn the clicks but to move the target: being named in the answer is the position that still converts.

Why a monthly brief exists: citations rotate constantly

The most underreported number in AI search is not traffic share. It is churn. Citations are not positions you hold, they are placements that rotate.

Corroborating evidence from every direction: AirOps found only about 30% of brands stay visible from one AI answer to the next for the same query. ChatGPT's own citation volume fell more than 90% at a March and April trough before rebounding in May, which reframes the trend as instability rather than decline. And brands are roughly 6.5x more likely to be cited through third-party sources than through their own domains, which is why earned media is the only structural defense against drift. Rankings don't pay you. Revenue does, and in a channel this volatile, revenue comes from consistent presence rather than a single won placement. The freshness mechanics behind this are in our citation freshness brief.

Money is arriving faster than expertise

The AI visibility category matured from a fringe idea into a funded software market and a budget line inside enterprise marketing.

Two consequences for buyers. First, tooling is consolidating around a split between purpose-built AEO platforms and AI modules bolted onto legacy SEO suites, and the honest answer on which to buy depends on company size, budget, and whether a team exists to act on the data. A dashboard nobody acts on is an expensive screenshot. Second, budget is moving faster than capability, which means a lot of spend is landing on measurement rather than on the earned media, content structure, and entity work that actually moves citations. That gap is the opportunity for brands that execute now.

What to do before the next cycle

Five moves that follow directly from what changed in July. This is the answer engine optimization (AEO) operating rhythm Lureon runs for clients across ChatGPT, Perplexity, Claude, Gemini, Grok, and Google AI Overviews.

One myth to retire while you plan: Google has stated that llms.txt files receive no special treatment in Search, including its generative AI features. Original research, first-hand expertise, and clean technical accessibility are what the guidance actually rewards.

What the operating rhythm produces

Lureon runs this as a continuous program rather than a one-time audit, which is the only structure that survives monthly citation turnover.

Lureon's work is led by a senior operator with 12+ years in search, everyone else on the team has 6+ years, no juniors, no outsourcing. For SaaS companies, scope lives on the B2B SaaS AEO service page and pricing is public.

Frequently asked questions