126 prompts · 6 use cases · 4 answer engines · 504 responses

AEO Wrangler B2B Software Index · October 2026

Executive summary

We asked ChatGPT, Gemini, Google AI Mode, and Microsoft Copilot 126 buying questions about simulation and analysis software (CAE). The questions spanned six use cases, from early design exploration to scaling simulation across an engineering organization. No prompt named a vendor or a product. Every name in this report comes from what the engines said. None came from what we asked.

Across 504 responses, the engines named 203 distinct software products. Of those, 174 are simulation and analysis products from 76 companies and a group of open source projects. The engines cited 3,944 URLs from 453 domains.

Eleven findings stand out.

Three companies own the answer. Ansys (Synopsys) scores 90.1 on equal channel weight, Siemens 84.5, and Dassault Systèmes 71.0. The fourth-placed company, SimScale, scores 32.7, a drop of 38.3 points. Together the three leaders account for 55.7 percent of all company mentions.

SimScale, the youngest company in the field, outranks incumbents decades older. Founded in 2012, SimScale is the only company in the top ten founded after 1990. It ranks fourth at 32.7, ahead of Autodesk, Cadence, PTC, and COMSOL, all founded in the 1980s. For companies under 100 people it scores 67.4, within 2.7 points of Siemens. On CFD questions it places third at 52.3, ahead of Dassault Systèmes. Its own site backs 72.7 percent of its mentions, the highest of any company in the top ten, a sign that the engines find SimScale through retrieval.

The product race at the top is close. Abaqus (41.3), Ansys Fluent (40.9), Ansys Mechanical (39.3), and Simcenter STAR-CCM+ (38.1) sit within 3.2 points of each other. Each leads a different slice of the market.

Structures and fluids are two different markets. On FEA questions, Dassault leads with 94.4 and Abaqus is named in 74.5 percent of prompts per engine. On CFD questions, Ansys and Siemens split the lead, and SimScale (52.3) and the open source projects (48.6) both outrank Dassault.

Company size changes the shortlist. For companies under 100 people, SimScale comes within reach of the leaders, Autodesk rises to 54.2, and open source to 44.4. For enterprises, Cadence climbs to fourth at 41.0.

The engines describe a market that no longer exists. Products that changed owner between 2024 and 2026 account for 37.8 percent of all product mentions. The engines rarely say so. Ansys products are named in 465 answers and Synopsys, their owner since July 2025, in 3. Former Altair products are named in 226 answers, and 182 of them give no sign that Siemens owns them. MSC products moved to Cadence in February 2026, yet 33 of the 75 answers naming them credit Hexagon and never mention Cadence.

Nastran is everywhere and belongs to no one. Nastran is named in 140 answers. In 51 of them, the engine never says whose Nastran it means. Those mentions credit neither Cadence, Siemens, nor Autodesk.

ChatGPT and Copilot read vendor sites. Google’s engines read everyone else. 90.9 percent of ChatGPT’s citations and 88.9 percent of Copilot’s point to vendor-owned domains. For AI Mode the figure is 32.3 percent, and YouTube alone supplies 13.0 percent of its citations. When ChatGPT names a simulation company, it cites that company’s own site 82.1 percent of the time. Gemini does so 17.5 percent of the time.

Dassault is the leader the engines remember. Dassault’s own sites back 9.7 percent of its Gemini mentions and 18.2 percent of its AI Mode mentions, the lowest of the three leaders. Abaqus is supported by a Dassault citation in 17 percent of its mentions on Google’s engines. When ChatGPT and Copilot need Abaqus documentation, they cite a university mirror of the Abaqus 2025 manuals 24 times and Dassault’s own help site twice.

No one owns the independent buying guide. Google’s engines draw about a fifth of their citations from “best software” and comparison pages. Of the 287 such citations in the study, independent blogs wrote 132 and software vendors only 22. A single blog page on the best CFD software was cited 52 times across 39 prompts. Trade media and analyst sites wrote 12, and 10 of those came from one publication, Demystifying PLM, whose 2026 guide to simulation software was cited 9 times across 8 prompts.

On AI questions, the incumbents set the agenda. Five prompts ask directly about machine learning and AI agents in simulation. Ansys SimAI is named in 15 of the 20 answers and Siemens’s Simcenter PhysicsAI in 13. Neural Concept, named in 11, is the one AI-native independent the engines consistently put beside them.

1. Why this index exists, and what it does not measure

Every simulation vendor’s marketing team is now asking the same question: when an engineer asks an AI assistant which FEA or CFD package to buy, does our name come up? Until now the answers have been anecdotal, based on a prompt someone tried once and a screenshot.

This index replaces that with a fixed, disclosed, repeatable instrument.

What it measures is visibility: whether a product is named, how often, by which engine, and in which buying situation. It is a share of voice study.

It does not measure solver accuracy, fit, or market position. Being named often is no evidence that a product is good, and being absent is no evidence that a product is bad or losing customers. Read every low placement in this report as a statement about AI visibility and nothing else.

This edition adds one check on the engines themselves. Section 10 asks whether the engines know who owns the products they recommend. In a category reshaped by five major acquisitions in two years, that is the fact a buyer most needs and the one the engines most often get wrong.

Editorial independence. AEO Wrangler has taken no vendor money in this category, has no paid placements in this index, and offers none. No vendor saw this report or its methodology before publication.

2. Method

Category definition. The core category follows CIMdata’s definition of the simulation and analysis (S&A) market. It covers structural analysis (linear, nonlinear, explicit, crash), structural dynamics and acoustics, multibody dynamics, CFD and thermal, electromagnetics, 0D and 1D systems simulation, simulation process and data management (SPDM), design optimization and design of experiments, pre- and post-processing, materials and manufacturing process simulation, and niche physics such as fatigue, fracture, and test correlation. General-purpose computation platforms such as MATLAB and Simulink, and multiphysics platforms such as COMSOL, are core.

Core and adjacent. The boundary was drawn after collection, from the data. A product is adjacent if its primary job is something other than simulation: CAD, PLM, cloud and HPC infrastructure, requirements management, or general tools such as Excel and Python. Cloud HPC platforms built to run simulation, such as Rescale, are treated as adjacent infrastructure.

Simulation sold under a CAD brand. Several vendors sell their simulation packages under the CAD product’s name: SOLIDWORKS Simulation, Fusion Simulation, Creo Simulation, Onshape Simulation, and Inventor’s stress and Nastran tools. Answers often recommend these by the CAD name alone (“use Fusion”, “SOLIDWORKS has this built in”). Mentions of SOLIDWORKS, Fusion, Creo, Onshape, and Inventor therefore count as mentions of the simulation package sold under that name.

Companies and products. The company ranking and the product ranking are kept separate throughout. A company’s score is the union of its products, so a prompt counts once however many of the company’s products it names. Acquired products roll up to the current owner, with the acquisition date recorded in Appendix D. Every ownership fact was verified against current sources in October 2026. Ansys is listed as Ansys (Synopsys), because Synopsys owns it and the engines know it as Ansys. Former Altair products are listed under their current Simcenter names.

Two reporting rows that are not companies. Open source projects such as OpenFOAM, SU2, and CalculiX have no vendor. Each appears as its own product in the product ranking, and the group appears as “Open Source” in the company ranking. When an engine names Nastran without saying whose, the mention credits no company and no product. Those mentions appear in the company ranking as “Nastran (vendor not stated)”, so the volume is visible without being assigned to anyone.

Use cases. Six, chosen from how CIMdata and practitioners describe the use of simulation across the product lifecycle.

Prompts. 21 per use case, 126 in total, in three categories.

High Buyer Intent (54 prompts) phrases an active purchase decision. Example: “I’m a mechanical engineer at a 40-person company that designs aluminum brackets and enclosures. I want to check stress and deflection on early concepts myself before we detail anything. What structural simulation software should we buy?”

Category With Intent (54 prompts) phrases a requirement that only a product can satisfy. Example: “I need browser-based fluid flow simulation that needs no local installation or hardware purchase and is affordable for a team of five, to check pressure drop on concept manifolds. What meets that?”

Open Question (18 prompts) phrases the problem as how or what approach, with a request for the tools involved. Example: “When a part fails a physical test, which simulation software should a team use to find and verify a fix?”

Physics and size. Within each use case, 9 prompts describe a structural (FEA) problem, 9 a fluid or thermal (CFD) problem, and 3 open questions name no physics. The 108 non-open prompts are split evenly across three company size bands by headcount: small (under 100), midsize (100 to 999), and enterprise (1,000 or more), 36 prompts each. No prompt names a vendor, product, or brand. One prompt names a standards body (ASME).

Engines and dates. ChatGPT (GPT-5.6 Sol, High), Gemini (3.6 Flash Extended), Google AI Mode, and Microsoft Copilot (Think Deeper, temporary chat). Copilot was used through its web interface in an incognito Chrome browser, with no connection to Windows or Microsoft 365. Responses were collected between September 25 and September 30, 2026. One Gemini response that was captured incorrectly was re-collected on October 11. Every prompt was run in a fresh incognito or temporary session on each engine, so no answer could be influenced by a prior prompt, account history, or personalization. Each response was captured in full with its source list.

Pilot and revision. The prompt set was piloted on ChatGPT first. Thirteen open questions that drew no product names were rewritten to ask for the tools involved, and ChatGPT was re-run on them before the other three engines were collected. All four engines answered the same final 126 prompts.

Extraction. Products were identified with a lexicon of alias patterns built by reading the responses, with no predefined vendor list. Engine citation labels and link targets were removed before matching, so a cited domain never counts as a mention. Sources were extracted as URLs, deduplicated within each answer, reduced to registrable domains, and classified by type. Citations are counted from each engine’s source list. For three Gemini answers whose sources appeared only as links inside the answer, those links were used.

3. The weighting decision

The four engines produce very different volumes of evidence. Copilot returned 10.2 citations per prompt, AI Mode 10.0, ChatGPT 8.7, and Gemini 2.3. Gemini gave no sources at all for 19 of 126 prompts, and AI Mode for 8.

That asymmetry does not distort the mention counts, because each engine answered the same 126 prompts and named a comparable number of products: 10.5 software products per answer for Gemini, 9.1 for ChatGPT, 8.5 for AI Mode, and 7.0 for Copilot. Source data is less balanced, so every source figure in this report is expressed as a share of that engine’s own citations.

The ranking uses equal channel weight: each company’s or product’s coverage rate within each engine, averaged across the four. Every engine contributes one quarter, however much it says.

We publish no usage-weighted score. Reliable usage shares for these four engines do not exist in a form we could defend. AI Mode is part of Google Search, and Copilot reaches users through Windows, Microsoft 365, and the web, so published usage figures for both cannot be separated cleanly from the products around them. This study used the web version of Copilot only. Every table reports per-engine rates beside the composite, so readers who prefer a different weighting can recompute the ranking themselves. Union coverage, the number of prompts where a company or product was named on any engine, appears as a reference column.

Inclusion floor. A company or product must be named in at least 8 of the 126 prompts (6 percent) to be ranked. Below that, the difference between two appearances and one is noise, so those entries are reported as an unordered set in Appendix A. Ties are broken by union coverage, then alphabetically.

4. Core simulation and analysis ranking

4.1 By company

The primary leaderboard. The equal weight score sets the rank: it is the average of the four engine coverage rates. Each engine column shows the percentage of the 126 prompts in which the company was named on that engine.

Below this line, 62 further companies were named in fewer than eight prompts.

Reading the table. The structure is three companies, then a cliff. Ansys and Siemens are named on more than 78 percent of prompts on every engine, and both are named somewhere in 122 of the 126 prompts. Their lead rests on breadth: each has a leading product in structures, in fluids, in optimization, and in simulation data management. Dassault reaches third through Abaqus on the structural side and the SIMULIA and SOLIDWORKS families around it, but its position depends on the engine. It is named in 81.7 percent of Gemini prompts and 54.0 percent of Copilot prompts.

The gap between third and fourth is 38.3 points. Below the leaders, the field divides into specialists that own a slice of the market. SimScale owns cloud CFD and small companies. Autodesk and PTC own simulation embedded in CAD. Cadence owns the enterprise and the MSC heritage. COMSOL owns multiphysics. HBK owns durability, and ESTECO owns optimization. Keysight, which owns the former ESI Group crash, casting, and acoustics portfolio, is named in 2 prompts.

4.2 By product

The same measure applied to individual products. Siemens places six products in the top twenty and Ansys five.

59 products clear the inclusion floor. The remainder appear in Appendix A. Per-product detail for the multi-product companies is in Appendix C.

The top four products sit within 3.2 points of each other and divide the market cleanly. Abaqus and Ansys Mechanical are the structural answers, Fluent and STAR-CCM+ the fluids answers. SimScale is the one product in the top five from an independent company, and OpenFOAM the one product with no vendor at all.

SOLIDWORKS Simulation’s place reflects the rule in section 2. Many answers name SOLIDWORKS as the environment in which a simulation tool runs, and those mentions count toward SOLIDWORKS Simulation. The same applies to Creo, Fusion, and Inventor. Without that rule, SOLIDWORKS Simulation would score 16.5.

4.3 SimScale: the newcomer among the incumbents

The top of this market is old. Ansys was founded in 1970, Dassault Systèmes in 1981, Autodesk in 1982, PTC in 1985, COMSOL in 1986, and Cadence in 1988. Siemens’s simulation business was assembled from companies with similar histories. SimScale was founded in 2012 by graduates of the Technical University of Munich, and it sells simulation through a web browser with nothing to install.

In AI answers it already outranks most of those incumbents.

How SimScale earns that position matters as much as the position. Several incumbents, Dassault above all, are named largely from memory. SimScale is named because the engines find it: when ChatGPT or Copilot names SimScale, they cite SimScale’s own pages 98 percent of the time, and on Gemini and AI Mode 58 percent of the time, against 17 to 37 percent for the four leading products. Its product pages and documentation give the retrieval engines current material to work with. A newer company can build that kind of visibility deliberately, and SimScale shows how far it reaches.

The gaps are equally clear. SimScale is named in 23.0 percent of ChatGPT prompts and 23.8 percent of Copilot prompts, against 44.4 percent on AI Mode, and it falls out of the top six for enterprises.

5. Extended ranking including adjacent software

The same measure applied to every software product named, whatever its category. Simulation products account for 84.9 percent of all software mentions. CAD accounts for 5.6 percent, PLM for 4.6 percent, and cloud and HPC infrastructure for 2.3 percent.

The CAD system frames the answer. Engines describe simulation tools in terms of the CAD environment they run in. Siemens NX is named in 66 prompts, as many as the leading simulation products, and CATIA in 47. Dassault’s 3DEXPERIENCE platform and Siemens’s Teamcenter are the PLM names the engines reach for when a prompt mentions traceability or simulation data management. Even so, simulation products account for more than four in five of all software mentions, so the answers stay focused on the simulation decision.

6. Per-engine leaderboards

The four engines agree on the top three companies. Below the top three, and in how strongly they favor each of the three, they disagree.

6.1 ChatGPT (GPT-5.6 Sol, High)

ChatGPT is the one engine that puts Siemens first, and the company it names first in an answer is Siemens 38.9 percent of the time. It is also Cadence’s strongest engine. It reads vendor documentation closely: 90.9 percent of its citations point to vendor-owned sites, 28.5 percent to documentation and support pages, and three of its 1,100 citations to comparison lists.

6.2 Gemini (3.6 Flash Extended)

Gemini names the most products per answer and shows the least evidence for them: 2.3 citations per prompt, with 19 of 126 answers carrying no sources. It is Dassault’s strongest engine at 81.7 percent, and it names Ansys Discovery in 34.9 percent of prompts, more than twice the rate of any other engine. Because Gemini cites so little, its rankings are the closest available view of what the model absorbed in training.

6.3 Google AI Mode

AI Mode is the retrieval engine of this study and the most open to influence. It cited 314 domains, more than twice as many as any other engine, and only 32.3 percent of its citations were vendor-owned. YouTube supplied 165 of its 1,266 citations, and community sites such as Reddit and LinkedIn another 9.6 percent. It names Ansys first in half of its answers. SimScale reaches its best figure here, 44.4 percent.

6.4 Microsoft Copilot

Copilot gives the shortest answers, with a median of 308 words, and names the fewest products: 6.0 simulation products per answer. Five of its answers named no simulation product at all, and 14 named two companies or fewer. When it does name a company, it cites that company’s own site 88.9 percent of the time, the highest support rate in the study. Dassault’s 54.0 percent here is its weakest figure on any engine.

6.5 Where the engines disagree

The gap between the two Google engines and the average of ChatGPT and Copilot, for ranked products with the largest gaps in each direction.

The engines that read vendor documentation favor specialist and enterprise tools such as AcuSolve, Fidelity, and COMSOL. Google’s engines favor the tools an engineer reaches most easily: CAD-embedded simulation, browser-based CFD, and Ansys Discovery. On this evidence, a buyer’s shortlist depends heavily on which assistant they open.

7. Use case, physics, and size

7.1 Use case leaderboards

Twenty-one prompts per use case, ranked by equal channel weight within the use case. Each table shows the top six. Companies whose use case rank differs from their overall rank by four or more places are marked with an asterisk.

Early Design Exploration (ED)

The one use case where the field opens up. Ansys Discovery (69.0) is the leading product, followed by SOLIDWORKS Simulation (57.1), SimScale and Creo Simulation (48.8 each), and Fusion Simulation (33.3). This is where simulation embedded in CAD, and simulation that runs in a browser, win the answer.

Design Verification (DV)

The most conservative use case. Sign-off questions draw the established solvers: Ansys Mechanical (53.6), Fluent (48.8), Abaqus (46.4), and STAR-CCM+ (44.0). Cadence reaches its best use-case score here, through MSC Nastran and its aerospace certification heritage.

Virtual Testing (VT)

Abaqus leads the products at 50.0, ahead of Fluent (46.4) and STAR-CCM+ (41.7). OpenFOAM (28.6) carries the open source group to fourth. HBK, through nCode, scores 17.9, its best use-case figure apart from troubleshooting.

Design Optimization (DO)

The one use case where Ansys does not rank first. The leading products are the optimization layers, Ansys optiSLang (52.4) and Simcenter HEEDS (46.4). ESTECO, with modeFRONTIER, scores 20.2 in eighth place, its best figure anywhere.

Troubleshooting and Failure Investigation (TF)

Ansys’s highest score anywhere: it is named in all 21 prompts on ChatGPT and Copilot. Ansys Mechanical (63.1) and Abaqus (53.6) lead the products. COMSOL (29.8) and HBK (26.2) follow in seventh and eighth, both at their best use-case figures.

Scaling Simulation Across the Team (ST)

SimScale ties Abaqus as the leading product here (44.0 each), ahead of STAR-CCM+ (35.7) and Fluent (34.5). Ansys Minerva (27.4) is the leading simulation data management product. These answers also name the most infrastructure: Rescale, AWS, Azure, and HPC schedulers appear more often here than anywhere else.

7.2 Physics: structures and fluids

On structures, Dassault leads. Abaqus is named in 74.5 percent of FEA prompts per engine and Ansys Mechanical in 66.2 percent. Gemini named Dassault in every one of the 54 FEA prompts.

On fluids, the field is wider. Fluent (82.4) and STAR-CCM+ (76.4) are the default answers, then SimScale (52.3) and OpenFOAM (48.1). CFD is where an independent company and an open source project both outrank one of the big three.

The 18 open questions name no physics. COMSOL scores 48.6 on them, more than three times its overall figure, and MathWorks 25.0. When a buyer asks a general question, the engines reach for multiphysics and systems tools.

7.3 Company size

Company size reshapes the field more than in any other dimension. For small companies, SimScale and Autodesk come within reach of the leaders, and open source scores 44.4. For enterprises, Siemens moves to first and Cadence to a clear fourth, while SimScale falls out of the top six.

8. Consolidation: what the engines know about who owns what

Five transactions between 2024 and 2026 changed the ownership of a large part of this market: Keysight’s acquisition of ESI Group, Cadence’s acquisition of BETA CAE, Siemens’s acquisition of Altair, Synopsys’s acquisition of Ansys, and Cadence’s acquisition of Hexagon’s Design and Engineering business, which includes MSC Software. Appendix D lists the dates.

A sixth is pending. On October 5, 2026, after this study’s responses were collected, Schneider Electric announced an agreement to acquire PTC, with closing expected by the third quarter of 2027, subject to shareholder and regulatory approval. Until it closes, this index credits Creo Simulation and Onshape Simulation to PTC. If it completes, Creo and Onshape will sit under a parent the engines do not yet associate with simulation, widening the gap between the product names buyers see in answers and the company that owns them.

We grouped every ranked and unranked product by how it reached its current owner.

Percentages are each cohort’s share of simulation product mentions on that engine.

Most of what the engines recommend was bought. More than two thirds of product mentions are for products that reached their current owner through an acquisition, and 37.8 percent for products whose owner changed within the last two years. Products built by their current owner account for 11.1 percent.

The acquisitions add visibility unevenly. For Siemens, the former Altair products (OptiStruct, HyperMesh, Inspire, SimSolid, Radioss, AcuSolve, and others) are named at an equal weight of 44.8, against 72.6 for Siemens’s other simulation products. Together they lift Siemens close to Ansys. For Cadence, the former MSC products score 14.9 and BETA CAE’s products 2.0, against 9.9 for Fidelity, Celsius, and Reality DC. Without the MSC products, Cadence would score 11.3, so the MSC heritage more than doubles its visibility. Even so, MSC Nastran (6.9) is outranked by Simcenter Nastran (9.9). Keysight’s former ESI portfolio is nearly invisible, named in 2 prompts.

Independents win the answer in a few places. 63 independent companies were named, and independent products appear in 80 to 100 prompts per engine. Most are named once or twice. SimScale, COMSOL, HBK, ESTECO, MathWorks, SimFlow, ENGYS, and Neural Concept are the independents that clear the floor, and each owns a specific situation: cloud CFD, multiphysics, durability, optimization, systems modeling, OpenFOAM-based CFD, and AI surrogates.

On AI questions, incumbents set the agenda. Five prompts ask about machine learning surrogates, AI-assisted optimization, or AI agents for simulation. Across the 20 answers, Ansys SimAI is named in 15, Simcenter PhysicsAI in 13, and STAR-CCM+ in 12. Neural Concept, named in 11, is the one AI-native company the engines consistently place beside them, and it clears the overall inclusion floor at exactly 8 prompts. Monolith is named in 7 of the 20, PhysicsX in 2.

Open source holds a real place in fluids. OpenFOAM is the seventh-ranked product overall and is named in 48.1 percent of CFD prompts per engine. No other open source project reaches 2 points. Open source matters to the engines as a CFD alternative and almost nowhere else.

9. Source analysis: what drives visibility

9.1 How the engines cite

A citation is a unique URL within one answer. ChatGPT and Copilot concentrate on a small set of vendor domains. AI Mode spreads across 314 domains, most of them cited once.

9.2 What kinds of sources

Gemini draws 14.4 percent of its citations from resellers and consultancies, the value-added resellers that sell and support simulation software. Their buying guides reach buyers on Google’s engines and almost nowhere else. ChatGPT and Copilot cite standards bodies and regulators (ASME, NAFEMS, NASA, the FDA, the FAA) more than any non-vendor source, mostly on verification and certification questions.

9.3 The citation support rate

For every company mention, we checked whether the same engine cited that company’s own domain in the same answer. Pages on hexagon.com that describe MSC products count as Cadence-owned, because Cadence now owns those products.

By company, for the ranked companies:

Mention prompts count one per prompt per engine, so the maximum is 504. “n/a” means the company was not named on that engine.

Two mechanisms. A company can be visible because the model learned it or because content exists and gets retrieved. SimScale, SimFlow, and ESTECO sit at the top of the support table: when they are named, their own sites are usually cited, which means they are being found. HBK, PTC, and MathWorks sit at the bottom: they are named without supporting citations, which means they are being remembered. HBK’s figure has a specific cause. Ansys resells nCode as “Ansys nCode DesignLife”, so engines often name nCode and cite an Ansys page.

9.4 Memory or retrieval: which products the engines remember

The support rate read product by product, split between the engines that cite vendor pages (ChatGPT and Copilot) and the engines that cite widely (Gemini and AI Mode).

Mentions count one per prompt per engine. Support means the same answer cited a domain owned by the product’s company. Readings: 65 percent or more is retrieval-driven, 50 to 64 mixed, 35 to 49 leaning recall, below 35 recall. OpenFOAM’s own-domain citations are to openfoam.com and openfoam.org.

Siemens is the most findable company in the category. Of the eleven retrieval-driven products in the table, six are Siemens products, three are Ansys products, and the others are SimScale and Cadence Fidelity CFD. Siemens’s documentation, blogs, and product pages (siemens.com was cited 577 times across 116 prompts) give every engine something to retrieve.

Dassault’s products are carried by memory. Abaqus is the most-named product in the study, and on Google’s engines its own company’s site supports 17 percent of its mentions. SOLIDWORKS Simulation, SOLIDWORKS Flow Simulation, and fe-safe show the same pattern. Dassault’s simulation documentation sits largely behind a login. When ChatGPT and Copilot need Abaqus documentation, they cite a university-hosted copy of the Abaqus 2025 manuals 24 times across 13 prompts, and Dassault’s own help site twice.

Memory is a strong position and a decaying one. Training data refreshes on a cycle no vendor controls. Retrieval indexes update continuously. A product whose visibility rests on recall has no defense against a competitor that out-publishes it in the retrieval layer, and the retrieval engines are where SimScale, SimFlow, and the CFD independents are concentrating their content.

9.5 What kinds of pages get cited

The engines split in two. ChatGPT and Copilot cite product, documentation, and pricing pages and almost never a “best of” list. Gemini and AI Mode draw about a fifth of their citations from listicles and comparison articles.

9.6 Who writes the lists

Of the 287 listicle and comparison citations in the dataset, 281 come from Gemini and AI Mode.

The most-cited lists:

The lists are written by independents, and few of them carry editorial standards. Software vendors wrote 22 of the 287 list citations. Most come from engineering blogs, training sites, and companies outside the category. One blog post on the best CFD software was cited in 39 of the 126 prompts. The one trade publication on the list is Demystifying PLM, whose guide to simulation software was cited on Gemini and AI Mode. With its companion guide to CAD software, it accounts for 10 of the 12 list citations from trade media and analysts.

9.7 Who owns the citation layer

Four observations for anyone trying to influence these systems.

Ansys now publishes from two domains. ansys.com and ansys.synopsys.com together appear in 120 prompts, more than any other company’s domains. The split has no visible cost today.

Altair’s domain still works for Siemens. altair.com was cited 251 times in 79 prompts. It is a Siemens asset now, and it is one of the main reasons the former Altair products are retrieved.

Hexagon’s pages still carry MSC. hexagon.com was cited 41 times, mostly for MSC products that Cadence now owns. Engines that read those pages repeat the old owner.

Trade media has left the most influential slot in the category largely open. Trade media and analyst sites account for 1.0 percent of all citations and 12 of the 287 list citations. Demystifying PLM is the most-cited trade or analyst domain in the study, with 10 citations, ahead of engineering.com with 6. Its 2026 simulation software guide is the one trade or analyst list cited more than once, and it shows that a structured, current comparison from a trade publication gets picked up. Most of the independent comparisons that Gemini and AI Mode read are still written by blogs and training sites. A trade title that published current, structured comparisons of FEA and CFD tools, with license models, host CAD support, and ownership stated plainly, would compete for those citations with a credibility advantage.

10. The buyer’s view: which engine to trust

Everything above treats the engines as a channel vendors want to be seen in. Buyers see the same answers from the other side. When an assistant states a fact, is it right? When it gives a shortlist, is the shortlist complete?

10.1 Ownership accuracy

Prices are hard to check in this category, because most simulation software is quoted on request. Ownership is easy to check and matters to a buyer: it decides who sells, supports, and funds the roadmap of the product being bought. We checked every answer that named a product affected by the 2024 to 2026 acquisitions.

Ansys is still Ansys. Synopsys completed its acquisition of Ansys in July 2025. The engines name Synopsys in 3 of 465 answers. For a buyer this matters less than the other two cases, because the Ansys brand remains in use.

Altair’s products are named as if Altair were independent. Siemens completed its acquisition of Altair in March 2025 and rebranded the products as Simcenter in June 2026. Across 226 answers that name former Altair products, the engines use the old Altair names 406 times and the current Simcenter names 73 times. 182 of those answers give no sign that Siemens owns the products. Several list Siemens and Altair as competing vendors.

MSC is credited to a company that sold it. Cadence completed its acquisition of Hexagon’s Design and Engineering business, including MSC Nastran, Marc, Adams, Apex, Cradle CFD, Actran, and SimManager, in February 2026. Gemini credits these products to Hexagon in 14 of 21 answers and never mentions Cadence. ChatGPT and AI Mode credit Hexagon without mentioning Cadence in roughly two answers in five. Copilot is the most current engine: 2 of its 13 answers credit Hexagon, and 5 name Cadence. ChatGPT is inconsistent within its own answers: in some it explains that the MSC portfolio moved to Cadence, and in others it describes “Hexagon SimManager” as a current product.

Smaller errors follow the same pattern. All three answers naming PV Elite describe it as a Hexagon product. Hexagon spun it off with Octave in May 2026, and one of the three answers mentions Octave. AI Mode describes SimScale’s structural solver as “Hexagon’s Marc” in three answers.

10.2 Completeness and depth

ChatGPT gives the most complete answer. It writes the longest answers, names close to eight simulation products in each, backs most of its company mentions with the company’s own pages, and recommends with reasons. Its ownership record is mixed.

Copilot is the most current and the narrowest. It is the engine most likely to know who owns MSC and to use the Simcenter names for former Altair products, and it backs nearly nine company mentions in ten with a vendor source. It also gives the shortest answers and names the fewest products. In one case it read a concept-stage structural analysis question as a building-engineering question and answered with civil structural tools.

AI Mode gives a broad view built from other people’s lists and videos. It cites more domains than any other engine. About a sixth of its citations are comparison lists and about an eighth are YouTube videos. It is useful for discovering names, especially cloud and open source CFD tools, and weak on evidence.

Gemini is the engine we would not rely on for a purchase decision. It names the most products and shows the least evidence for them. A sixth of its answers cite nothing, it cites the named company’s own site for 17.5 percent of company mentions, and it is the most out of date on ownership.

Our ranking for buyers in this category: ChatGPT first, Copilot second, AI Mode third, Gemini fourth. ChatGPT combines depth, breadth, and evidence. Copilot matches its evidence and leads on ownership accuracy with a fraction of its breadth. AI Mode offers breadth with secondhand evidence. Gemini offers breadth with little evidence and the most out-of-date picture of who owns what.

10.3 What this means for a buyer

Start with ChatGPT, and check its shortlist against a second engine. The shortlist depends on the engine: SOLIDWORKS Simulation appears in 47.6 percent of AI Mode answers and 14.3 percent of Copilot’s, and SimScale in 44.4 percent of AI Mode’s and 23.0 percent of ChatGPT’s. If you use Copilot, ask it explicitly for alternatives.

Confirm who owns the product before you buy it. The engines regularly name the former owner of Nastran, Marc, Adams, SimManager, OptiStruct, HyperMesh, and their neighbors. The current owner is the company that will quote, support, and develop the product.

When an answer names Nastran, ask whose. MSC Nastran, Simcenter Nastran, and Inventor Nastran are different products from three different companies.

Ask for sources, and look at them. When the source is a “best software” list, check who wrote it. In this study most of those lists came from blogs and training sites with no stated testing method.

11. What this means

For Ansys (Synopsys). The visibility contest is won at company level. Ansys leads on three of four engines, is named first in a third of all answers, and leads or ties five of six use cases. Its product lead is narrower: Abaqus edges Fluent and Mechanical, and STAR-CCM+ is within three points. The exposure is on Google’s engines, where Ansys’s own sites support 12.4 percent of its Gemini mentions. The Synopsys name is almost absent from the answers, which carries no cost while the Ansys brand stays in use.

For Siemens. Siemens is first on ChatGPT, first for enterprises and design optimization, and the most findable company in the category. The Altair acquisition is working in visibility terms: the former Altair products add reach across structures, optimization, and durability. The open issue is naming. The engines still call these products Altair and often present Altair as a competitor. altair.com remains one of the most-cited domains in the study, and until it and the third-party pages it feeds reflect the Simcenter names, the engines will keep repeating the old ones.

For Dassault Systèmes. Dassault owns structures. Abaqus is the most-named product in the study and Dassault leads every FEA measure. That position rests on memory. On Google’s engines, Dassault’s own pages support fewer than one Abaqus mention in five, and the engines that read documentation are citing a university copy of the Abaqus manuals. Copilot names Dassault in 54.0 percent of prompts, its weakest figure on any engine. Opening documentation to crawlers is the most direct lever available.

For SimScale. SimScale is the strongest result in this study for a company outside the incumbent group. Founded in 2012, it ranks fourth among companies, ahead of competitors with three or four decades more history, and it is the default answer for small companies and cloud CFD. Its visibility is retrieval-driven, strongest on Google’s engines, and supported by its own pages more consistently than any leader’s. The next gain is ChatGPT and Copilot, where it is named in fewer than a quarter of prompts, and the enterprise buyer, where it drops out of the top six.

For Cadence. Cadence now owns one of the broadest simulation portfolios in the market, and the engines have not caught up. Its MSC products are often credited to Hexagon, its BETA CAE products are rarely named, and many Nastran mentions credit no one. ChatGPT and Copilot, the engines that read vendor pages, give Cadence its best figures. Moving MSC content from hexagon.com to Cadence-owned pages, and stating the ownership plainly, is the fastest fix available.

For Autodesk and PTC. Both are visible through simulation embedded in their CAD products, and both are strongest in early design. Autodesk is also a strong answer for small companies. Their visibility leans on recall: the engines name Fusion, Inventor, and Creo as places where simulation happens, and cite Autodesk and PTC pages for a minority of those mentions. PTC also faces a naming change: Schneider Electric’s pending acquisition, announced after this study was collected, will put Creo and Onshape under a new owner once it closes.

For the independents. COMSOL, HBK, ESTECO, MathWorks, SimFlow, and ENGYS each own a specific buying situation, and each is close to invisible outside it. The independents with the highest support rates (SimFlow, ESTECO, SimScale) are the ones being found. Those with the lowest (HBK, MathWorks) depend on being remembered.

For the AI-native companies. The engines answer AI questions with the incumbents’ AI products. Neural Concept is the exception, named beside Ansys SimAI and Simcenter PhysicsAI on most AI prompts. Every other AI-native company is named once or twice.

For Keysight. The former ESI Group crash, casting, and vibro-acoustic tools are close to absent. Two prompts named a Keysight product. On crash questions, the engines answer with LS-DYNA, Radioss, and Abaqus.

For trade media and independent reviewers. The engines are answering buying questions from comparison articles, and almost none of those articles come from publications with editorial standards. Demystifying PLM’s simulation guide is the exception, and it is cited in 8 prompts. A publication that owns the independent buying guide for FEA and CFD, with current ownership stated, would supply the evidence Gemini and AI Mode are currently taking from blogs.

For everyone. The engines describe a market that existed before 2025. A vendor whose ownership, product names, or documentation location changed in the last two years should assume the engines have not noticed, and should check.

12. Limitations

Collection window. Responses were captured between September 25 and September 30, 2026, with one Gemini response re-collected on October 11. These systems change continuously and a rerun would produce different numbers.

Single run per prompt per engine. No repeat sampling, so variance within an engine is not measured. Some of the long tail is likely noise.

Model version specificity. Results describe GPT-5.6 Sol High, Gemini 3.6 Flash Extended, Google AI Mode, and Copilot Think Deeper as they behaved in September 2026. Other tiers of the same products may behave differently.

CAD-branded simulation. Mentions of SOLIDWORKS, Fusion, Creo, Onshape, and Inventor count as mentions of the simulation package sold under that name, including answers that name the CAD product only as the environment another tool runs in. This raises the scores of those products and of Dassault, Autodesk, and PTC. Siemens NX and CATIA were not treated this way and remain adjacent.

Prompt interpretation. One Copilot answer read a concept-stage structural analysis prompt as a building-engineering question and named civil structural tools. Those tools are counted as named.

Lexicon extraction. Products were matched by curated alias patterns. A product named once under an unusual spelling could be missed. Products named in fewer than five prompts should be treated as directional.

Ownership. Ownership facts were verified against company announcements and current product pages in October 2026. The current owner of ODYSSEE, a product that came with Hexagon’s former Design and Engineering business, could not be confirmed. nCode is credited to HBK, its developer, although Ansys resells it.

Source classification. Domain types were classified by hand for the most-cited domains and by rule for the rest. Page types were classified from URL patterns. hexagon.com counts as Cadence-owned when supporting MSC product mentions.

Citation support is a proxy. A mention accompanied by a citation to the company’s own domain suggests retrieval-driven visibility. It does not prove causation.

No prompt volume data. This study measures how engines answer these questions. It does not measure how often real buyers ask them.

Appendix A: Companies named in fewer than eight prompts

Reported as an unordered set: Advanced Design Technology, AirShaper, Aras, Arup (Oasys), AVL, Bentley Systems, Bramble, CFD Support, CFturbo, Codeware, Collier Aerospace, Computers and Structures Inc., Convergent Science, CoreTech System (Moldex3D), DATADVANCE (pSeven), DHCAE Tools, Diabatix, Dive Solutions, Dlubal, Dynamic Design Solutions (FEMtools), Flexcompute, Flow Science, Fracture Analysis Consultants, Friendship Systems (CAESES), FunctionBay, Gleason (KISSsoft), Gradient Dynamics, HTRI, Ideal Simulations, Inter-CAD, Karamba3D, Keysight, Khorium, Luminary Cloud, Magna (FEMFAT), Maya HTT, Mecway, MESYS, Monolith AI, Nemetschek, Numeric Systems, Octave, Paulin Research Group, PhysicsX, Prometech, Sente Software (JMatPro), Shonan Design, Simerics, SkyCiv, SmartUQ, SOFiSTiK, Southwest Research Institute (NASGRO), Synera, Tecplot, Thermo-Calc Software, ToffeeX, Trimble, Wolf Star Technologies, Wolfram, Zentech, d3VIEW, nTop.

Appendix B: Adjacent software by category

CAD (6): Siemens NX, CATIA, Solid Edge, Fibersim, Rhino and Grasshopper, FreeCAD.

PLM (4): 3DEXPERIENCE platform, Teamcenter, Windchill, Aras Innovator.

Cloud and HPC infrastructure (8): AWS, Microsoft Azure, Google Cloud, HPC schedulers (Slurm, PBS, LSF), Rescale, Simcenter HPCWorks, Simr, TotalCAE.

Other adjacent (11): Excel, Python, NVIDIA GPUs and frameworks, Codebeamer and other requirements tools, Revit, Speckle and iTwin, Hexagon Nexus, PTC ThingWorx, Oqton 3DXpert and Materialise, Wallacei, Altium and Keysight ADS.

Appendix C: Multi-product companies, per-product detail

Products above the inclusion floor.

Combined figures are the union of prompts naming any of the company’s named products. They are smaller than the sum of the rows.

Appendix D: Ownership changes reflected in this index

Pending: Schneider Electric’s agreement to acquire PTC, announced October 5, 2026, with closing expected by the third quarter of 2027. PTC products are credited to PTC in this index.

Appendix E: Prompt set and underlying data

The complete set of 126 prompts, the six use case definitions, the locked product lexicon, and the full mention and source datasets are available at no charge on request to info@aeowrangler.ai.

AEO Wrangler is an independent consultancy specializing in Answer Engine Optimization. This index is editorially independent and carries no paid placement.

Source: AEO Wrangler, “The Simulation and Analysis Software AI Visibility Index,” AEO Wrangler B2B Software Index, October 2026.