Daily SEO Note — September 8, 2026: Apple Says Applebot-Extended Rules Do Not Touch Search Ranking

SEO for Content Writers

One item cleared the evidence bar in the last 24 hours. Apple's Applebot documentation, dated September 4, 2026 and picked up across the SEO press on September 7, now states in plain language that site rules for Applebot-Extended are not considered in ranking for Search. That moves the AI-training opt-out out of the risk column and into the same drawer as a syndication or licensing decision — which is where it should have been sitting all along.

Apple States the Applebot-Extended Opt-Out Is Not a Ranking Factor

What changed: Apple's About Applebot support page now carries the sentence "Site rules for Applebot-Extended are not considered in ranking for Search." Applebot-Extended has only ever governed whether content Applebot already crawled may be used to train Apple's general-purpose foundation models; it never controlled search inclusion. What was missing until now was Apple saying the ranking part out loud, and that absence is what kept the decision parked in a lot of editorial calendars. The page is dated September 4, 2026, and the status is live documentation rather than a phased rollout.

Who it affects: all content, on every site Apple crawls — but most acutely publishers whose differentiator is original reporting, proprietary benchmarks, or first-hand testing, and who have been deferring the opt-out because nobody could promise it was free. Google made the equivalent commitment for Google-Extended some time ago: disallowing it does not affect inclusion in Google Search and is not a ranking signal. Both major training opt-outs now carry the same written guarantee.

What to do differently: treat the training opt-out as a rights and licensing question, and move it out of the SEO section of your brief entirely. If your original data is the asset that earns citations, you can now decline training use without hedging the recommendation. What to stop doing: stop describing an Applebot-Extended block as a traffic risk in stakeholder memos, and stop bundling it with Applebot itself. Blocking Applebot does remove you from Apple's search surfaces. Blocking only Applebot-Extended does not.

Google Reiterates That Low-Value Programmatic Pages Cost Site-Level Trust

On September 7, 2026, Google Search Advocate John Mueller said publicly that programmatic SEO often produces pages that are spam, borderline spam, or simply low quality, and that when it does, Google's systems can lose faith in the site as a whole rather than in the offending pages alone. Treat this as a spokesperson restatement, not a documented system change — the policy of record remains the scaled content abuse section of Google's spam policies, which is unchanged. Nothing shipped; the guidance simply got sharper about where the damage lands.

Who it affects: any content operation running templated pages at scale — location grids, "X vs Y" comparison matrices, auto-generated category and integration pages. It does not target templates as a technique. It targets templates whose only variable is a substituted entity name.

What to do differently: cap the number of near-identical variants per cluster, and require one non-templated element on every page in it — a price checked this week, a first-hand observation, a screenshot from an actual test, a named source. What to stop doing: stop treating the cleanup of a decayed programmatic cluster as a page-level fix scheduled around other work. By Mueller's account the trust loss is site-wide and slow to reverse, which makes pruning a priority rather than a backlog item.

No Ranking or Spam Update Is In Flight

The Google Search Status Dashboard showed no crawling, indexing, ranking, or serving incidents as of September 7, 2026 at 23:02 PDT, and the Search Central documentation changelog has published nothing since August 31 — the last confirmed ranking update remains the August 2026 spam update.

Apply to Your Next Brief

• Move the Applebot-Extended and Google-Extended decision from the SEO section of the brief to the rights and licensing section. Neither affects ranking, and both vendors now say so in writing.
• Write the training opt-out and the search crawler as two separate lines in every policy note. Applebot and Googlebot stay allowed; only the -Extended agents are in question.
• For any templated cluster, name in the brief the one first-hand element each page must carry. If you cannot name it, the page does not get commissioned.
• Add a pruning line to this quarter's plan for any cluster whose pages differ only by a swapped entity name.
• Do not brief around a live core or spam update this week. None is running, so volatility you see is not an announced event.
• Keep unconfirmed SERP sightings out of client-facing recommendations until a primary source carries them.

SEO for Developers

No framework, browser, spec, or security release landed inside the window. The only work worth a same-day pull request today is crawler policy, and there are two concrete edits: Apple's documentation change removes the last reason to leave Applebot-Extended undecided in robots.txt, and OpenAI's bot documentation has moved to a new host and now publishes an IP range file for a fourth crawler that most robots.txt files have never had a rule for.

robots.txt: Split the Training Agents From the Search Agents

Identifier and date: Apple's About Applebot support document, last updated September 4, 2026. Non-breaking. There is no technical symptom if you ignore it; the cost is a policy decision left permanently open because nobody could quantify the ranking risk. That risk is now documented as zero.

File to change: public/robots.txt. Write four distinct groups so the search crawler and the training agent are never governed by the same rule on either vendor.

# Apple: Applebot controls Search inclusion.
# Applebot-Extended controls only whether crawled content trains
# Apple foundation models. Its rules are not used in Search ranking.

User-agent: Applebot
Allow: /

User-agent: Applebot-Extended
Disallow: /

# Google: same split. Google-Extended is training-only and is not
# a ranking signal, and does not affect inclusion in Google Search.

User-agent: Googlebot
Allow: /

User-agent: Google-Extended
Disallow: /

One parsing trap worth naming: under RFC 9309, a crawler obeys the group whose user-agent line matches it most specifically, and consecutive user-agent lines form a single group sharing one rule set. Do not stack Applebot and Applebot-Extended as adjacent user-agent lines above one Disallow — that is a single group, and it will pull your site out of Apple's search index rather than out of its training corpus. Blank-line separation between groups is what keeps the two decisions independent.

OpenAI's Crawler Docs Moved, and OAI-AdsBot Now Publishes IP Ranges

Identifier and date: the legacy documentation path platform.openai.com/docs/bots now returns HTTP 301 to developers.openai.com/api/docs/bots, verified today. Non-breaking, but any runbook, README, or compliance document that hard-codes the old path is now one hop stale. The page documents four agents, each a separate decision: OAI-SearchBot/1.4 for ChatGPT search surfacing, GPTBot/1.4 for foundation-model training, OAI-AdsBot/1.0 for validating pages submitted as ChatGPT ads, and ChatGPT-User/1.0 for user-triggered fetches.

The change worth acting on is OAI-AdsBot. When it first appeared in the docs there was no published IP range file for it, which made verification impossible; openai.com/adsbot.json now exists and carries a creationTime of 2026-05-12. Symptom if ignored: a blanket disallow on a marketing host, or an edge rule that blocks unrecognised agents, can stop ad landing-page validation and leave campaigns unapproved for reasons that never surface in the ad platform. The mirror-image mistake is blocking OAI-SearchBot while intending only to opt out of training — that removes you from ChatGPT search citations and leaves the training door exactly where it was.

# OpenAI ships four agents. Four separate decisions.

# Search surfacing in ChatGPT — allow to stay citable.
User-agent: OAI-SearchBot
Allow: /

# Foundation-model training — disallow to opt out.
User-agent: GPTBot
Disallow: /

# Ad landing-page review — must be reachable or ads fail approval.
User-agent: OAI-AdsBot
Allow: /

# User-triggered fetch on a person's behalf.
User-agent: ChatGPT-User
Allow: /

Enforce it at the edge on IP, not on the user-agent string, which anyone can forge. OpenAI publishes searchbot.json, gptbot.json, adsbot.json, and chatgpt-user.json under openai.com; pull all four into whatever allowlist your WAF or bot rules already use, and apply the same discipline Google documents for verifying Googlebot. If your CDN sits in front of robots.txt, confirm it is serving your file rather than a cached or platform-injected variant — a bot rule that contradicts your robots.txt intent is the failure mode nobody notices until citations disappear.

Quiet Window Everywhere Else

Checked and empty for the window: no GitHub advisories on September 7 or 8 affecting Next.js, Nuxt, Astro, SvelteKit, or any sitemap and SEO package; Next.js stable still on the 16.3 line with 16.4 in canary and nothing metadata- or sitemap-related in it; Schema.org still on release 30.0 from March 19, 2026; and no Core Web Vitals threshold or measurement change.

Ship Today

1. Rewrite public/robots.txt into four separated groups — Applebot, Applebot-Extended, Googlebot, Google-Extended — and record the training decision in the repo rather than in someone's memory.
2. Add an explicit OAI-AdsBot group on any host that serves ad landing pages, then confirm no WAF or bot rule blocks it above robots.txt.
3. Confirm OAI-SearchBot is allowed if ChatGPT search citations are wanted. It is a different line from GPTBot and the two are routinely conflated.
4. Refresh the crawler verification allowlist from all four OpenAI IP range files; adsbot.json will be absent from any config built before May 2026.
5. Update hard-coded references to platform.openai.com/docs/bots in runbooks and compliance docs to the developers.openai.com path.