Why September 2026 Matters
September 2026 produced six concrete signals that reshape how brands should approach AI search visibility. They range from a model change inside Google AI Mode, to a discovery about how ChatGPT makes brand recommendations, to data showing AI-referred traffic is growing faster than most predictions. Together they suggest the "AI search" conversation is splitting into two distinct disciplines, and most teams still treat them as one.
1. Gemini 3.7 Flash Lands in AI Mode
Google rolled Gemini 3.7 Flash into AI Mode one day after the model's public release, available to paid subscribers through the model picker. Google describes the update as "better at following instructions and understanding intent" (GPO, September 14, 2026, citing Google).
This follows Gemini 3.5 Flash in July, then 3.7 in September. The cadence is the new variable: AI Mode citations should now be tracked across model updates, the way traditional SEO is tracked across algorithm updates. A baseline this month may not hold next month.
For teams: log the model id (when visible) and capture AI Mode citation presence per priority query. A shift between 3.5 and 3.7 may be a model effect rather than a content signal.
2. ChatGPT Recommendations Rely on Training, Not Live Web
Testing of ChatGPT's recommendation behavior for category-level queries (e.g., "what's the best brand for X") found the model leans heavily on associations formed during training, not live retrieval. Brands not already connected to their category in training data had roughly a 2% chance of being recommended (GPO, September 14, 2026, citing testing).
This is the most strategically loaded finding of the month. Most "GEO" work targets citation: getting your page parsed and surfaced as a source. That work matters for factual queries. But for category-level recommendations, the model's prior knowledge dominates, and the prior comes from sustained brand presence across the web: earned media, PR, third-party coverage.
The implication: a brand with no third-party footprint cannot optimize its way into ChatGPT recommendations. The fix is brand-building, not technical SEO.
3. ChatGPT's Fetch Bot May Ignore robots.txt
OpenAI clarified in September 2026 that ChatGPT-User, the bot that fetches a page live when a user references a specific URL, may not honor robots.txt disallow rules. TollBit's H1 2026 data found roughly 15% of AI page fetches on tracked European sites reached disallowed pages (GPO, September 14, 2026, citing TollBit).
Anthropic, by contrast, says all three of its bots honor robots.txt consistently.
For teams: robots.txt is still useful for crawlers that respect it, but it is not a reliable barrier against a user-initiated fetch. Genuinely private content (drafts, internal tools, staging) needs to sit behind authentication, not rely on a directive the fetcher may ignore.
4. AI-Referred Traffic Grew 197% Year-Over-Year, and It Converts
Shopify's Q2 2026 data found AI-referred storefront sessions grew 197% year-over-year, converting at roughly twice the rate of organic in research-heavy categories. Organic search still drove more total traffic than all AI platforms combined (GPO, September 14, 2026, citing Shopify Q2 data).
Two reads follow:
- AI traffic is no longer a rounding error. It is growing faster than organic and converting well in specific intent bands.
- The AI-versus-organic framing is the wrong frame. Both are growing, and the investments that benefit one largely benefit the other.
For most brands we work with, the practical takeaway is resource allocation. AI-search work does not compete with organic SEO. It shares the content substrate. The same structured product data, clean schema, and answer-first copy make both channels stronger.
5. ChatGPT Indexes Roughly 200 Characters of Body Content
Analysis of ChatGPT's free-tier index found it stores only a page's title and approximately 200 characters of body content, about two sentences. Any boilerplate, generic introduction, or navigation text before the first substantive paragraph consumes that space (GPO, September 14, 2026, citing index analysis).
| Template issue | Common effect | Why it matters |
|---|---|---|
| Hero brand statement before substantive copy | Eats the first 100-150 characters | The "AI index budget" is spent on branding, not the answer |
| Disclaimer or cookie banner text adjacent | Compiled text source includes this | Cited text in an AI response may pull boilerplate |
| Generic intro paragraph | Same boilerplate across pages | Pages lose distinctiveness in the AI retrieval layer |
The fix: pages where AI citation matters (service pages, location pages, product pages) should put the first substantive sentence as early in the body as possible. Templates that open with brand statements are spending the most valuable index budget on text that won't get cited.
6. Google's SerpApi Lawsuit Could Tighten Rank Tracking Data
Google refiled its DMCA lawsuit against SerpApi after an initial dismissal, narrowing the claim to licensed content from partners like Reddit who required Google to block unauthorized scraping (GPO, September 14, 2026).
SerpApi's data feeds a significant share of rank-tracking and SERP feature reporting tools the SEO industry runs on. Even a narrow Google win could narrow access to scraped SERP data and raise reporting tool costs.
No immediate action needed, but worth knowing which tools in your stack rely on scraped SERP data. First-party data from Google Search Console and GA4 becomes more valuable if third-party access tightens.
What the Six Signals Mean Together
September splits the "AI search" conversation into two distinct problems:
Citation: a technical problem
Getting a model to find, parse, and surface your page as a source. Addressed with crawlability, content structure, schema, and placing substantive copy at the top of the page. Fast feedback loop. You can see whether you are being cited and iterate.
Recommendation: a brand problem
Getting a model to recommend your brand for category-level queries. Addressed with earned media, PR, third-party coverage, and sustained presence across the web the models were trained on. Long feedback loop. Models shift over months.
Most teams are over-invested in the first and under-invested in the second. September's data suggests that imbalance is now the strategic risk.
For SMBs specifically: technical citation optimization is necessary but cheap. The recommendation work, building the brand web that the models remember, is harder, slower, and more aligned with how brands have always earned trust. Worth investing in early.
The One Takeaway
Citation and recommendation are not the same problem. Citation is an optimization problem with a short feedback loop. Recommendation is a brand-building problem with a long feedback loop. A team investing only in the first will keep ranking for queries it already ranks for, while losing ground on the queries where AI actually names a brand.
Read next:
- Shopify's "Slop Grenades" Warning and What It Means for Marketing Teams
- Alibaba's AgentCore and Agentic Cloud: What the Full-Stack AI Strategy Means
- Google's AI Threat Report: Agents Steal Thousands of Credentials in 6 Hours
- SEO to AEO: A Practical AI Search Playbook
Want to see how AI search actually sees your brand, citation and recommendation both? Contact Vivid for an AI visibility audit that separates the two problems.
