LinkedIn Articles vs Posts: AI Citations Fell 65% in Six Months

LinkedIn is routinely described as the second most cited domain in AI search. We run a monitoring corpus that can check that claim directly, so we did.

Does AI cite LinkedIn articles?

Yes, and much less than it did at the start of 2026. Across 390,151 AI answers collected in the 30 days to 7 August 2026, linkedin.com was the 14th most cited domain, appearing in 1.61% of answers. YouTube led at 14.07% and Reddit followed at 11.30%. On a fixed prompt set measured monthly, LinkedIn’s share of answers fell by 65% between February and July 2026, and the Article surface specifically fell by 74%.

Articles carry almost all of what is left: they appear in 59.1% of LinkedIn-citing answers against 33.3% for feed posts and 4.2% for company pages. On the same corpus, a listicle on your own domain is cited more often than every LinkedIn Article on the internet combined.

LinkedIn AI search citations fell 65% in six months

A share of a growing corpus can fall without anything actually changing. Our monitoring corpus grew roughly 18 times between February and July 2026 as customers added brands and prompts, so any domain’s share could fall simply because the mix of questions changed.

To remove that, we held the questions still. The table below uses only the 116 prompts that ran in every one of the six months, restricted to the four conversational engines (ChatGPT, Perplexity, Google AI Mode, Copilot). Same prompts, same engines, same brands, six consecutive months.

Month	Answers	Any LinkedIn URL	LinkedIn Articles
February 2026	12,393	13.80%	10.73%
March 2026	13,520	9.02%	6.46%
April 2026	29,878	7.42%	4.64%
May 2026	13,260	5.94%	4.06%
June 2026	12,753	5.54%	3.65%
July 2026	11,991	4.83%	2.84%

The takeaway: the decline is monotonic across all six months and nothing about the measurement changed between them, so this is a real, sustained trend to plan around, not a one-month blip.

One denominator check matters here. If engines simply cited fewer sources overall, every domain’s share would fall together. They did cite fewer. Average sources per answer went from 13.95 in February to 11.19 in July, a 20% drop, while the share of answers carrying any source at all stayed between 92% and 99%. A 20% thinning cannot produce a 65% fall, and as the next section shows, it did not apply evenly.

This is not a LinkedIn problem, it’s a shift in what counts as a citable source

Neither, which is the more interesting answer. On the identical 116-prompt cohort, the platforms moved in different directions.

Platform	February 2026	July 2026	Change
Reddit	30.43%	18.18%	down 40%
Wikipedia	17.22%	1.32%	down 92%
LinkedIn	13.80%	4.83%	down 65%
YouTube	11.74%	16.31%	up 39%
Medium	4.24%	2.47%	down 42%

YouTube gained share while the text platforms lost it, and Wikipedia fell hardest of all. This is a rebalancing of what engines treat as a citable source, not a penalty aimed at LinkedIn. LinkedIn fell faster than Reddit and Medium and slower than Wikipedia.

That also updates advice we published ourselves. Our post on Reddit’s citation decline suggested professional platforms such as LinkedIn as a place to rebuild. On this data, LinkedIn was falling on the same curve, roughly one year later.

Which AI engines cite LinkedIn? The spread is more than a hundredfold

Averages hide almost everything here. The spread across engines is more than a hundredfold, measured over the 30 days to 7 August 2026.

Engine	Answers	Any LinkedIn	LinkedIn Articles
Google AI Mode	53,636	6.20%	4.07%
Google AI Overview	56,364	3.97%	2.18%
Perplexity	55,611	2.29%	1.27%
Copilot	55,479	0.67%	0.54%
ChatGPT	56,276	0.57%	0.06%
Google News	56,362	0.08%	0.05%
Gemini	56,423	0.00%	0.00%

The Google AI Overview and Google News rows count sources present on the results page, which includes organic results alongside AI Overview citations, so they are not strictly comparable to the conversational engines. The four chat engines are directly comparable, and they are the four used in the trend above.

Gemini cited a LinkedIn URL once in 56,423 answers. That is not missing data. Gemini carried at least one source on 55.8% of its answers at 4.4 sources each, and on the same corpus it cited Reddit in 3.53% of answers and YouTube in 2.95%. It cites user-generated content; it does not cite LinkedIn. Our earlier work on grounding by engine covers why Gemini’s source behaviour differs from the rest.

Copilot is the only engine that cites LinkedIn more than Reddit. Copilot cited Reddit in 0.00% of answers and YouTube in 0.78%, against 0.67% for LinkedIn. Microsoft owns both Copilot and LinkedIn. We can measure the pattern; we cannot see the ranking system that produces it, so treat the ownership as context rather than cause.

LinkedIn articles vs posts: articles win, company pages barely register

The usual version of this question is about reach and engagement, which citation data cannot settle. What it does settle is which of the two gets quoted. Articles appear in 59.1% of LinkedIn-citing answers whereas feed posts appear in 33.3%, and company pages barely register at 4.2%. Of the 7,580 answers that cited any LinkedIn URL in the 30-day window:

LinkedIn surface	Share of LinkedIn-citing answers
Article (/pulse/)	59.1%
Feed post (/posts/)	33.3%
Company page (/company/)	4.2%
Showcase page	0.8%
Personal profile (/in/)	0.5%
LinkedIn Learning	0.1%
Jobs	0.1%
Other (topic hubs, /advice/ collaborative articles)	7.6%

Shares are of answers, not of citations, and they total more than 100% because 5.8% of these answers cite several kinds of LinkedIn URL at once.

The takeaway: most brands concentrate their LinkedIn effort on the company page, which is the surface AI engines quote least, and on feed posts, which vanish from the feed in days. The long-form Article is the surface carrying almost 60% of LinkedIn’s remaining citation weight.

One caution against over-investing. The window holds 5,119 Article citations spread across 1,550 distinct URLs, and 62.6% of those URLs were cited exactly once. Citations spread thin rather than concentrating on a few evergreen Articles.

LinkedIn Article versus an owned listicle: it depends on the engine

This is the comparison that decides where the next hour of writing goes. Across all 390,151 answers in the window, each tied to a tracked brand:

Surface	Share of answers citing it
Any page on the brand’s own domain	19.86%
Any listicle, on any domain	43.15%
A listicle on the brand’s own domain	1.44%
Any LinkedIn Article	1.15%

An owned listicle is cited slightly more often than every LinkedIn Article on the internet put together. The brand’s own domain, across all page types, is cited seventeen times more often than LinkedIn Articles.

The largest number in the table is the one that is not about you at all. Third-party listicles appear in 43% of AI answers. Getting into other people’s roundups outweighs both owned and LinkedIn publishing by an order of magnitude.

By engine, the choice inverts:

Engine	LinkedIn Article	Owned listicle
Google AI Mode	4.07%	1.39%
Google AI Overview	2.18%	1.13%
Perplexity	1.27%	2.01%
Copilot	0.54%	1.16%
ChatGPT	0.06%	3.47%
Gemini	0.00%	0.34%

If ChatGPT is where your buyers ask, a LinkedIn Article is close to worthless and your own comparison page is worth roughly 58 times more. If Google AI Mode is the surface that matters, LinkedIn Articles outperform your own listicle by three to one. There is no single answer, and any advice that gives one is averaging across engines that behave in opposite ways.

We classify a listicle by URL pattern (best-, top-N, -vs-, alternatives, -review, buying-guide, comparison). Reading a sample of 25 matched URLs by hand, 24 were genuine roundups or comparisons. The classifier runs a little loose, so read the listicle figures as close estimates rather than exact counts.

Why the “second most cited domain” figure is stale

The claim that LinkedIn is the second most cited domain in AI search, at roughly 11% of responses, traces to OtterlyAI’s study of January to June 2026.

Our data does not contradict it so much as date it. Our February 2026 reading on the fixed cohort was 13.8%, in the same region as their figure for a window centred earlier in the year. By July the same measurement returned 4.83%, and over the most recent 30 days LinkedIn sits 14th among all domains at 1.61%.

The study was reasonable for the period it covered. Citation share moved faster than the studies describing it, which is the general hazard with any AI-visibility benchmark quoted more than a quarter after it was run. Where our data does agree with Otterly is on the shape. Citation behaviour is uneven by engine: Copilot leans almost entirely on Articles (81% of its LinkedIn citations are /pulse/ URLs), and Gemini barely cites LinkedIn at all.

What to do with LinkedIn now

LinkedIn still earns citations on the Google surfaces, and it costs nothing to publish there. The measured position supports a modest, specific role rather than a channel strategy.

1. On LinkedIn articles vs posts, write the article. Articles carry 59.1% of LinkedIn’s citations, feed posts 33.3%, company pages 4.2%.
2. Pick the surface your buyers actually use. Google AI Mode cites LinkedIn Articles at 4.07% and ChatGPT at 0.06%. That single fact decides whether the effort is worth it.
3. Do not move budget from owned content. An owned listicle already outperforms all LinkedIn Articles combined, and the brand’s own domain outperforms them seventeenfold.
4. Spend the larger share on third-party roundups. Being in other people’s listicles reaches 43% of answers, more than every owned and LinkedIn surface in this study put together.
5. Re-measure quarterly. A figure from two quarters ago described a platform with three times the citation share it has now.

How we measured this

The corpus is cloro’s AI monitoring database: 1,531,027 AI answers collected between 22 November 2025 and 8 August 2026, carrying 12,046,775 recorded source citations. The engines are ChatGPT, Perplexity, Google AI Mode, Google AI Overview, Copilot, Gemini and Google News. Metrics use only answers with a SUCCESS status.

Two windows appear above. Current figures use the 30 days to 7 August 2026 (390,151 answers). Trend figures use the 116 prompts present in all six months from February to July 2026, restricted to the four conversational engines, so the question set is constant across the series.

“Cited in X% of answers” always means the share of answers carrying at least one source from that domain, never the share of citation slots. An answer citing a domain five times counts once.

Three limits worth stating plainly:

The prompt mix is commercial

These are customer brand-visibility prompts, weighted toward retail, travel, consumer goods and software. A corpus of B2B or recruiting questions would very likely place LinkedIn higher. The trend holds the mix constant, so the direction is sound within this corpus; the absolute levels are specific to it.

Country subdomains are counted

de.linkedin.com and its siblings added roughly 1,300 answers on top of the 6,296 for linkedin.com in the 30-day window. Domain rankings use linkedin.com alone; LinkedIn surface breakdowns include all subdomains.

We measure citations, not traffic or conversions

Being cited is not the same as being clicked, and this study says nothing about what any of these surfaces earns downstream.

If you want the same measurement on your own brand and prompts rather than ours, that is what the AI visibility tracking API exists to do.