We respect your privacy.

We use strictly necessary cookies to keep you signed in and to protect against CSRF. With your permission we also use a small amount of first-party analytics to improve the product. We do not sell your data and we do not use third-party advertising trackers. See our cookie policy and privacy policy .

← All posts

The ten blue links are back as a tab

Crawlmind Engineering··5 min read

A web-only view is a results mode that strips the AI answer and other features off a search page and shows a plain list of links to web pages, and all three major answer surfaces now either offer one or are testing one.

Google has had its version the longest. The "Web" filter launched in May 2024 and shows only text links. AI Overviews do not appear inside it. In mid-November 2026, Search Engine Roundtable reported that Bing is testing a "Web Mode" tab that removes the Copilot answers and the other features, leaving web page results. The same day it reported that OpenAI appears to be testing web results layouts in ChatGPT: a web-only list, web results shown before the richer answer, and an AI answer with a "See more" button above the links.

Treat the Bing and ChatGPT items as sightings, not launches. Neither company announced anything, the Bing tab did not look fully live to the reporter, and the author of the ChatGPT piece wrote that he could not replicate the variations himself. The direction is still worth planning for, because it changes what "visible" means on each surface.

#Two surfaces on one query

An AI answer and a list of links reward different things. The answer is assembled from passages. A page can be useful to it because one paragraph states a fact cleanly, even if the page as a whole is not the best result for the query. The link list ranks whole pages, and the user picks one by reading a title and a snippet.

That means a single query now has two scoreboards on the same engine. A brand can be cited in the AI answer and sit far down the web list, or rank first in the web list and be absent from the answer. A lot of AI visibility reporting collapses this into one number taken from the answer layer only. When a meaningful share of users switch to the links view, that number describes less of what they see.

Nobody has published how many people use these views. Google has not released usage of its Web filter, and the Bing and ChatGPT versions are tests. So the honest position is that the split is unknown, and the cheapest response is to measure both layers instead of guessing which one matters more.

The two layers are separate in presentation, not in plumbing. Google's documentation says a page has to be indexed and eligible to show with a snippet to appear as a supporting link in AI Overviews or AI Mode, with no extra technical requirements. The same crawl, index and snippet pipeline feeds both the answer and the plain list. A page that is blocked, noindexed or snippet-restricted loses both.

OpenAI's crawler documentation draws a sharper line. It says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers but can still appear as navigational links. If ChatGPT grows a web-results list, that distinction becomes visible to users. A publisher that blocked OAI-SearchBot to stay out of answers could still find its homepage in a link list, without the summary or the citation context. That may be exactly what the publisher wants. It is worth knowing which state you chose on purpose.

#What changes in practice

None of this undoes the work that earns AI citations. Clear, factual passages, accurate structured data and crawlable pages help both layers. The changes are mostly in measurement and in a few page elements that the answer layer ignores but the link layer depends on.

Titles and snippets matter again on AI-heavy queries. In a plain list, the title and the snippet are the whole pitch. Pages written for passage extraction sometimes carry generic titles because nobody expected a person to read them. Check the titles and meta descriptions on pages that win citations, and make sure each one says what the page is and who it is for.

Snippet controls cut both ways. Google lists nosnippet, data-nosnippet and max-snippet as the ways to limit what appears in AI features. They also shape the classic snippet in the web list. A restriction added to keep text out of AI answers can leave a page with a thin or empty snippet in the links view. Audit these directives before adding new ones and record why each exists.

Robots decisions need a link-layer column. For each AI crawler you allow or block, write down the expected result in the answer and in the link list. For OpenAI, blocking OAI-SearchBot removes you from answers according to its own documentation but not necessarily from navigational links. For Google, there is no separate switch: AI features and classic results come from the same index.

Rank tracking is not obsolete. Classic position tracking describes the web-only views directly. Keep it for the queries where AI answers appear, rather than retiring it because the answer took the top of the page.

#How to measure both layers

Start with the query set you already track for AI visibility. For each query, record two observations per engine on the same day:

  1. Whether your domain is cited in the AI answer, and which URL.
  2. Where your best URL ranks in the web-only view (Google's Web filter today, Bing's Web Mode where it appears).

Report the two series side by side, never averaged. The interesting rows are the disagreements. A page that ranks well but is never cited usually has the answer buried, phrased vaguely, or split across sections. A page that is cited but ranks poorly is often a strong passage on a weak page, and its title and snippet are the first things to fix.

Annotate the test dates. If Bing or ChatGPT ships a web-only view widely, mark the date on your charts so a change in click-through is not read as a ranking change. For ChatGPT, keep the layout you captured with each result, since the same prompt may return an answer-first page for one user and a links-first page for another while tests run.

Finally, keep checking your own referral data by landing page. If a links view sends traffic, it tends to land on the page that ranks, which is not always the page that gets cited. That difference is the clearest sign that both layers are now doing work, and that a single visibility score was hiding one of them.

Related field notes

Share or discuss

Field notes in your inbox

New posts, no spam. Roughly monthly. Unsubscribe with one click.