From Shubhi K | Product & Market Analysis

Software Developer Job Postings on Indeed: 18% Below 2020 and Rebounding

On this page

Software developer job postings on Indeed stood at 81.6 on 30 October 2026, where 100 is the level on 1 February 2020. That is 18.4% below the pre-pandemic baseline. All US postings on the same index sat at 103.8. The gap is real, it is narrowing, and most of it opened before any AI coding tool could plausibly have caused it.

Key takeaways

  • Software postings remain 18.4% below February 2020. Indeed's US software development index read 81.6 on 30 October 2026, while the all-postings index read 103.8 on the same date.
  • Most of the collapse predates ChatGPT. The index fell from 233.8 in February 2022 to 142.5 by 30 November 2022, which is 53% of its total fall to the May 2025 low of 61.1.
  • The rebound is real and narrow. Software postings rose 29% between Claude Code's preview on 24 February 2025 and 30 October 2026, while all postings fell about 5%. Hiring Lab attributes 71% of a year of gains to senior roles.
  • Postings are not hires. The index counts advertisements on one job site. It cannot show how many developers were hired, at what level, or whether a role was filled at all.
81.6US software development postings index on 30 October 2026, against 100 in February 2020. Source: Indeed Hiring Lab via FRED, retrieved November 2026.
+29%Change in software postings from Claude Code's preview to October 2026. All postings fell 4.8%. Source: Zan Digital calculation on Indeed data via FRED.
71%Share of the May 2025 to May 2026 increase in software postings that came from senior roles. Source: Indeed Hiring Lab, July 2026.

The latest software developer job postings reading

The most useful public check on claims about AI and engineering demand is a single daily series. Indeed Hiring Lab publishes it, and the St. Louis Fed republishes it on FRED as series IHLIDXUSTPSOFTDEVE. It tracks software development job postings in the US, indexed to 1 February 2020.

On 30 October 2026 the series read 81.6. In plain terms, there were about 18% fewer software postings on Indeed than in the last month before the pandemic. A year earlier, on 30 October 2025, the reading was 65.9, so the index rose about 24% in twelve months.

The aggregate index for all US postings read 103.8 on the same date. It rose about 3% over the same twelve months. So software is recovering faster than the wider market while still sitting far below it.

That framing matters because the two loudest stories about developer hiring both misread this chart. One says AI has already destroyed demand for engineers. The other says the 2026 rise proves AI is creating it. The data supports neither claim at full strength, and the rest of this post explains why.

Indeed postings index readings at dated points, US, 1 Feb 2020 = 100
DateEventSoftware developmentAll postingsSoftware as share of all
1 Feb 2020Baseline100.0100.0100.0
28 Feb 2022Software peak233.8160.4145.8
30 Nov 2022ChatGPT launch142.5146.797.2
14 Mar 2023GPT-4 release102.3135.275.7
24 Feb 2025Claude Code research preview63.2109.058.0
22 May 2025Claude Code generally available62.1106.058.6
30 Oct 2026Latest observation81.6103.878.6

Index values from FRED series IHLIDXUSTPSOFTDEVE and IHLIDXUS, retrieved 10 November 2026. The last column is Zan Digital's calculation: software index divided by all-postings index, times 100. Event dates are from the vendors' own announcements. Last verified 10 November 2026.

How the Indeed Hiring Lab software jobs index is built

Before you read anything into a number, you need to know what it counts. Hiring Lab documents the method in the public README for its job postings tracker. It is short, and three details shape every conclusion below.

A seven-day average, seasonally adjusted, against one day

Each value is the percentage change in seasonally adjusted postings since 1 February 2020, using a seven-day trailing average. A reading of 81.6 therefore means postings were 18.4% below that single February day. Hiring Lab seasonally adjusts each series separately, using a daily method developed by the Deutsche Bundesbank.

The baseline is one day, not a long average. That choice is reasonable for a pandemic tracker. It is less neutral for software, a category that was already growing quickly in early 2020. A different base would move every percentage in this post, though not the shape of the line.

Categories come from job titles

Hiring Lab assigns postings to occupational sectors using normalised job titles. Software development is an Indeed categorisation, not a government occupation code. A role titled "AI engineer" or "machine learning engineer" may or may not land in it, and the README does not list every mapping.

This is the detail most commentary skips. If employers retitle engineering roles around AI, some demand could migrate between categories without any change in the work. The index cannot separate a relabel from a real shift.

History gets revised

Hiring Lab adopted its current seasonal method in November 2024 and revised historical numbers, which may now differ from values reported earlier. You can see revision at work in its own posts. The July 2026 Hiring Lab analysis put software postings 27.5% below their pre-pandemic level. The current FRED vintage shows 73.5 for 1 July 2026, or 26.5% below. Part of that difference is dating and part may be revision. You cannot tell which from outside.

Software postings fell further, earlier, than the wider market Indeed US job postings index, seasonally adjusted, 1 Feb 2020 = 100. Selected dated observations. 200 150 100 50 0 ChatGPT, 30 Nov 2022 Claude Code preview, 24 Feb 2025 233.8 peak, Feb 2022 61.1 low, May 2025 81.6 All postings 103.8 Feb 2020 2022 2024 Oct 2026 Software development All postings Source: Indeed Hiring Lab via FRED, retrieved 10 Nov 2026
Look at where the dark line sits when the first red marker arrives. Software postings had already given back most of their 2021 boom before ChatGPT launched.

Software postings against overall postings

A single index can mislead because the whole labour market moves. Overall postings on Indeed rose to about 160 in early 2022 and then cooled for three years. Some of what happened to software is that cycle, and some is specific to software.

Dividing the software index by the all-postings index gives a cleaner signal. It asks how software is doing relative to everything else on the same site, measured the same way. This ratio is our calculation, not a Hiring Lab series, and the working is in the table above.

The ratio was 145.8 at the software peak in February 2022. By ChatGPT's launch it had already fallen to 97.2, roughly back to parity with the wider market. It then dropped to 75.7 by GPT-4's release in March 2023 and bottomed near 58 in early 2025. By 30 October 2026 it had recovered to 78.6.

I think the ratio is the more honest number to quote. The headline "18% below baseline" mixes software weakness with general softness. The ratio says software is about 21% weaker than the market as a whole, which is the claim people usually mean when they cite this chart.

Software postings relative to all postings on Indeed Software index divided by all-postings index, times 100. Parity with Feb 2020 mix = 100. 100 = parity 30 Nov 2022, ChatGPT97.2 14 Mar 2023, GPT-475.7 1 Jan 202460.3 24 Feb 2025, Claude Code58.0 22 May 2025, Code GA58.6 1 Jan 202665.2 30 Oct 202678.6
The relative trough sits around the Claude Code launches, not after ChatGPT. Whatever recovered after May 2025 did so while coding agents were spreading.

Developer hiring data around the AI milestones

The dated milestones give a crude but public test. If AI coding tools drove engineering demand down, the steepest declines should follow the tools. If something else drove it, the timing should not line up.

Most of the fall came before ChatGPT

Software postings peaked at 233.8 on 28 February 2022. By 30 November 2022, the day OpenAI launched ChatGPT, the index stood at 142.5. That is a fall of 91.3 points before any widely used chat model existed.

The index then fell another 81.4 points to its May 2025 low of 61.1. So 53% of the total decline happened before ChatGPT. The claim that AI killed software jobs does not survive that timeline. The end of the 2021 hiring boom is visible in the data well before the first credible culprit arrived.

This matches what employment data shows for the occupation. Our reading of the BLS occupational employment counts for software developers found the steepest occupational losses also predate the chat-model era, in a different dataset with a different method.

Where the software postings index lost its points Indeed US software development index, 1 Feb 2020 = 100. Change in index points. Baseline 100 233.8 Peak, Feb 2022 -91.3 Before ChatGPT -81.4 Nov 2022 to May 2025 +20.5 Rebound to Oct 2026 81.6 Latest, 30 Oct 2026 Source: Indeed Hiring Lab via FRED, retrieved 10 Nov 2026. Point changes are Zan Digital calculations.
The first red bar is bigger than the second. The rebound bar is small against both, which is why "recovery" overstates where the series actually is.

The trough came after Claude Code shipped

The second half of the decline is harder to explain away. Between GPT-4 in March 2023 and May 2025, software postings kept sliding while the wider market cooled more slowly. The relative ratio fell from 75.7 to about 58 in that window.

Anthropic released Claude Code as a limited research preview on 24 February 2025 and made it generally available on 22 May 2025. The software index hit its low of 61.1 on 17 May 2025, between those two dates. Then it turned. That is an awkward fact for the displacement story, because demand began recovering just as the most capable coding agents reached general release.

I would not read causation into either direction here. A single turning point near a product launch is a coincidence until something else confirms it. The defensible statement is narrower: the agent era has not coincided with a fresh collapse in posted demand.

What drove the 2026 rebound in software postings

Hiring Lab's own analysis, by economist Guillermo Gallacher in a July 2026 post on AI and job postings, found software postings had risen almost 15% since Claude Code's launch while overall postings fell 7%. On the current FRED vintage through 30 October 2026, that gap has widened to a 29% rise against a 4.8% fall.

Senior roles carried most of it

The same Hiring Lab post decomposed the increase between May 2025 and May 2026. It found that 71% of the rise in software development postings came from senior roles. The post does not publish how seniority is classified, so treat the split as directional.

That composition is the most important fact in the rebound. A recovery led by senior postings is consistent with teams adding experienced engineers to supervise and review AI-generated code. It is also consistent with employers simply rebuilding the senior bench they cut first. Our analysis of how the junior developer pipeline is breaking covers what a senior-only recovery means for the people trying to enter the field.

AI-titled roles are about a third of the gain

Hiring Lab also found that 37% of the increase came from jobs mentioning AI in their title. The senior and AI-titled groups overlap, so you cannot add the two shares. Some of those postings are probably new work. Some are probably existing engineering roles renamed for a market that rewards the label.

This is where I part company with the optimistic reading. A rebound that is heavily senior and heavily AI-titled could be the start of broad demand. It could equally be a narrow hiring pattern at well-funded AI product teams. The index cannot tell you which, and the honest move is to wait for another two or three quarters of data.

Percentages from a low base flatter. The index rose 33.5% from its May 2025 low, which sounds like a boom. In points, it climbed 20.5 from 61.1 to 81.6, which recovers less than a quarter of the 172.7 points lost from the peak. The low itself sat below the 64.7 reading the series touched on 19 May 2020, during the first pandemic lockdown.

What job postings cannot show about hires

Everything above is about advertisements. None of it is about people in jobs. That distinction is where most public use of this chart goes wrong, and it decides how much weight you should put on the rebound.

A posting can stay open for months, be reposted, be duplicated across boards, or never be filled. Employers sometimes post to build a pipeline for roles they have not approved. Others fill senior engineering roles through referrals and recruiters without posting at all. The index sees the first group and misses the second.

Hiring Lab's README is explicit that its numbers describe postings on Indeed for information purposes. It does not claim they measure hires, employment or vacancies across every channel. Treat any article that says "developer hiring is up 29%" from this series as having overstated its source.

What the postings series supports, and what you need instead
Claim you might want to makeDoes the Indeed index support it?Better evidence
Employers are advertising fewer software roles than in 2020Yes, for postings on IndeedThe index itself, with its baseline stated
Software demand is weak relative to the wider marketYes, via the ratio to all postingsThe ratio, plus other job boards for cross-checking
Fewer developers are employed than beforeNoBLS occupational employment estimates, which lag by a year or more
AI caused the declineNo, and the timing argues against it as the main causeFirm-level studies that compare AI adopters with non-adopters
Junior developers are being hired lessPartly, through seniority mix of postingsEmployment data by age, and graduate outcome surveys

The obvious fix is to check postings against official hires. It does not work cleanly. The BLS Job Openings and Labor Turnover Survey publishes openings and hires by industry, region and establishment size, not by occupation. There is no official monthly count of software developer hires to compare against.

That leaves annual occupational employment estimates and payroll-based studies, both of which arrive late. Our breakdown of what payroll data actually shows about AI job losses walks through what those sources can and cannot settle. The practical consequence is simple. For the next year at least, postings are the fastest signal you have, and the least complete.

Where this reading is weakest

The argument above leans on one company's data. Here is where it could be wrong.

One platform's view of the market

Indeed is large, but it is not the whole market. If software employers have shifted postings toward LinkedIn, company career sites or specialist boards since 2022, the Indeed series would understate demand. Nothing in the published method corrects for a platform losing share in one category. I have no data showing that shift happened, and I have none ruling it out.

The 2021 boom distorts every comparison

The 233.8 peak was an extraordinary moment of hiring, fuelled by pandemic demand for software. Measuring the decline from that peak makes it look catastrophic. Measuring from February 2020 makes it look serious but survivable. My 53% pre-ChatGPT figure depends on choosing the peak as the start point, and a reader who thinks the boom was the anomaly could fairly frame it differently.

The best case for the AI explanation

The displacement view has a real argument. The relative ratio kept falling from 97 to 58 after ChatGPT, while the wider market cooled. That post-2022 slide is software-specific, and AI tools are one candidate cause. Rising interest rates, the end of zero-cost capital for startups, and layoffs that corrected pandemic overhiring are others, covered in the analysis of AI layoffs as an overhiring correction. The posting data cannot rank them.

So the restrained reading is this. AI is not the main explanation for the fall from the peak. It may be part of the explanation for why software lagged the market after 2022. It has not, so far, coincided with a second collapse.

Frequently asked questions

Are software developer job postings back to pre-pandemic levels?

No. Indeed's US software development postings index read 81.6 on 30 October 2026, where 100 equals the level on 1 February 2020. That is about 18% below the pre-pandemic baseline. All US postings on Indeed stood at 103.8 on the same date, so software remains one of the weakest categories even after a 33% rise from its May 2025 low.

Did AI cause the drop in software job postings?

The timeline does not support AI as the main cause. Software postings peaked at 233.8 on 28 February 2022 and had already fallen to 142.5 by ChatGPT's launch on 30 November 2022. That pre-ChatGPT fall is 53% of the total decline to the May 2025 low. AI may have contributed after 2022, but postings data alone cannot isolate its effect.

How does Indeed Hiring Lab measure software jobs?

Hiring Lab assigns postings to occupational sectors using normalised job titles. It reports each sector as a seasonally adjusted, seven-day trailing average indexed to 1 February 2020. A reading of 81.6 means postings are 18.4% below that date. Each series is adjusted separately, and historical values were revised when the method changed in November 2024.

Do job postings mean companies are hiring developers?

Not directly. A posting is an advertisement that can stay open, be duplicated across boards, or never be filled. Indeed's index measures postings on Indeed, not hires, not employment and not every recruiting channel. To check hiring you need employment data such as BLS occupational estimates, which arrive with a lag of a year or more.

Why are software job postings rising in 2026?

Hiring Lab found that 71% of the increase in software development postings between May 2025 and May 2026 came from senior roles, and 37% came from jobs with AI in the title. Those groups overlap. The rise is real in the data, but it starts from a deep trough, so part of it is recovery rather than new demand.

How to use this chart in a hiring plan

If you set engineering headcount, the series is worth one bookmark and one habit. Open the FRED series on the first working day of each month and record two numbers: the software index and its ratio to all postings. The ratio is the one that tells you whether competition for engineers is tightening against the wider market.

Then do the thing the index cannot do for you. Pull your own last twelve months of engineering requisitions and split them by level. If your senior roles are taking longer to fill while junior applications pile up, your company is living inside the same 71% senior rebound Hiring Lab measured, and your 2027 plan should price senior hires accordingly.

Method note

Every index value in this post comes from the FRED vintage retrieved on 10 November 2026. Indeed revises its history, so a value you pull later may differ slightly. Ratios and point changes are our own arithmetic, shown in the tooltips.

References

  1. Indeed Hiring Lab via Federal Reserve Bank of St. Louis, Software Development Job Postings on Indeed in the United States (IHLIDXUSTPSOFTDEVE), daily, retrieved 10 November 2026. Used for every software index value.
  2. Indeed Hiring Lab via Federal Reserve Bank of St. Louis, Job Postings on Indeed in the United States (IHLIDXUS), daily, retrieved 10 November 2026. Used for all-postings values and ratios.
  3. Indeed Hiring Lab, job_postings_tracker README, GitHub, accessed November 2026. Used for index method, title-based categories and the November 2024 revision.
  4. Guillermo Gallacher, Indeed Hiring Lab, AI and Job Postings: From Destruction to Creation?, 8 July 2026. Used for the 15%, 7%, 27.5%, 71% and 37% figures.
  5. Indeed Hiring Lab, US Labor Market Snapshot, August 2026, 24 August 2026. Used to cross-check the 101.8 overall and 74.4 software readings for 14 August against FRED.
  6. Anthropic, Claude 3.7 Sonnet and Claude Code, 24 February 2025. Used for the research preview date.
  7. Anthropic, Introducing Claude 4, 22 May 2025. Used for the Claude Code general availability date.
  8. US Bureau of Labor Statistics, Job Openings and Labor Turnover Survey data, accessed November 2026. Used for the industry-level scope of JOLTS. ChatGPT (30 November 2022) and GPT-4 (14 March 2023) dates are from OpenAI's launch announcements.

The weakest point in this source base is that every postings figure comes from one platform, Indeed, with no independent cross-check from another job board. Seniority and AI-title shares rest on a Hiring Lab classification whose method is not published in the post. Last verified 10 November 2026.

SK
Sanskriti Khandelwal
Writes for Zan Digital about AI, work and the labour-market data behind claims about both.

Related reading