LinkedIn Posts Overtook Articles in Google’s AI Answers

GEOforge Research · Citation Data

All summer, Google’s AI answers cited LinkedIn articles far more often than LinkedIn posts. In the last week of August that flipped. Posts have out-cited articles every week since.

By the GEOforge TeamGEOforge6 min read

A simple question started this: when AI search cites LinkedIn, does it cite articles or posts? Over the last 90 days the answer is “articles, by about a third.” That answer is out of date. It averages two very different periods, and the more recent one points the other way.

What is the short answer?

• Until late August, Google’s AI answers favored LinkedIn articles. From mid-July to late August, articles were 63% of the LinkedIn articles and posts that Google AI Mode and AI Overviews cited, and they led every week. They led in June and early July too.
• From the week of 24 August, article citations fell by more than half and did not come back. Since 31 August, posts have out-cited articles every week. Articles are now 43% of the pair.
• It happened on both Google engines and in every brand with enough data. It is not one client’s topic mix moving the average.
• ChatGPT barely cites LinkedIn at all, and when it does, it cites posts.

What did we measure?

GEOforge runs each brand’s tracked prompts against ChatGPT, Google AI Mode and Google AI Overviews on a schedule and stores every source each answer cites. For this piece we used the 12 full weeks from 13 July to 4 October 2026: 292,958 answers to 786 prompts across 26 brands, citing 3,516,674 sources in total. 49,134 of those citations, about 1.4%, point to LinkedIn.

We sorted each LinkedIn link by its URL. Articles live under /pulse/. Posts live under /posts/ (or, rarely, /feed/update/). Company pages, personal profiles, LinkedIn’s own “top content” pages and videos make up the rest. Because the number of answers we collect varies from week to week, the trend charts use citations per 1,000 answers rather than raw counts.

When did articles lose their lead?

The week of 24 August. Before it, Google’s two AI surfaces cited LinkedIn articles between 107 and 167 times per 1,000 answers, and posts between 62 and 91. That week both dropped. Posts started climbing again the following week and reached 119 per 1,000 by the end of September. Articles did not: they have stayed between 37 and 47 per 1,000 answers ever since.

The drop was not a gradual drift. In the weeks before it, Google cited around 450 different LinkedIn articles a week. Since 31 August it has cited around 115. The pool of articles Google is willing to cite shrank by roughly three quarters, while the pool of posts stayed about the same size.

Did both Google engines change?

Yes. Comparing the six weeks before 24 August with the six weeks from it, article citations fell on both engines: by 75% in AI Mode and 41% in AI Overviews, per 1,000 answers. Posts fell much less in AI Mode and held up in AI Overviews.

Is this just one or two brands?

No. LinkedIn citations are concentrated: in our data, the two brands with the most LinkedIn article and post citations account for about half of them. So we checked brand by brand, comparing 3–23 August with 7–27 September. Every brand with at least 15 LinkedIn article or post citations in both windows saw articles lose share, nine out of nine.

These brands track different prompt sets on different topics. A shift that shows up in all of them points to a change in how Google picks LinkedIn sources, not a change in what our clients ask about.

What about ChatGPT?

ChatGPT is a different story, and a short one. Over the last 90 days it cited LinkedIn posts 134 times and LinkedIn articles 3 times. LinkedIn as a whole is about 0.31% of everything ChatGPT cites, roughly a fifth of its share in Google’s AI answers. That matches what we found in our study of 10.2 million AI citations: social platforms are mostly a Google phenomenon. If ChatGPT is where your buyers are, LinkedIn is a small lever whichever format you publish in.

What does LinkedIn citation look like overall now?

Putting all three engines and every kind of LinkedIn page together, the mix has moved in six weeks. Posts are now the largest category, and company pages and personal profiles have both gained share as articles shrank.

What should you do with this?

• If your LinkedIn plan for AI search rests on long-form articles, revisit it. Articles were the better bet for Google’s AI answers until August. Right now, posts are cited more often.
• Keep both formats and watch the trend, not the average. A 90-day average told us articles were ahead. The weekly data said the opposite. Six weeks is a real pattern, but it is still six weeks.
• Measure Google and ChatGPT separately. One blended LinkedIn number hides that ChatGPT barely uses LinkedIn while Google uses it a great deal.
• Treat this as “what Google cites”, not “why”. Our data shows which LinkedIn pages AI answers point to. It cannot tell you why Google changed, and we have not tested whether any particular kind of post earns more citations.

How did we measure this?

Data. Every answer GEOforge collected between 13 July and 4 October 2026 that came with a list of cited sources: 292,958 answers across 26 brands and 786 prompts, excluding Google answers collected on 27–28 August, whose source lists were Google redirect links. We counted the sources each engine attached to its answer. We did not count links that appear only in the answer text, because full-text scans over this volume were not feasible; a spot check of ChatGPT’s answer text over the last 30 days found 2 article links and 37 post links, in line with its source lists.

Classification. By URL path. A LinkedIn link under /pulse/ is an article and one under /posts/ or /feed/update/ is a post. Short links (lnkd.in) did not appear in Google’s source lists.

Normalization. Weekly collection volume varies, so trends are reported per 1,000 answers. The before and after periods are six full weeks each, split at the week of 24 August.

Checks. We ruled out three alternative explanations before writing this. A change in collection volume: weekly volume stayed between about 13,000 and 26,000 Google answers, and the charts are normalized per 1,000 answers anyway. A change in which brands were measured: the brand-level comparison holds in all nine brands with enough data. A change in our own code: nothing that parses Google’s source lists changed in that window.

Limitations. The prompts are the ones our clients track, so they lean B2B and are not a random sample of all searches. We cannot see why Google changed, and the halving of AI Mode’s source lists in the same week appears to come from Google. Every figure is aggregated; no brand or prompt is identified.

Sources & method. All figures come from the GEOforge production citation database: 13 July to 4 October 2026, 26 brands, 786 prompts, 292,958 AI answers and 3,516,674 cited sources, of which 49,134 point to LinkedIn. ChatGPT figures cover 8 July to 6 October 2026. June and early-July baselines from the same database. Engine-level social citation shares from GEOforge’s earlier study, Engagement Is a Threshold, Not a Ladder.