The 2026 French SMB digital report: web, SEO, GEO and business software, without the spin

Methodology and scope of the report

A report is only as good as the provenance of its numbers. This one covers French SMBs, meaning micro, small and mid-sized businesses from a handful of employees up to two hundred and fifty, excluding large accounts and venture-funded startups. It spans three areas that structure the digital life of a company: web presence and visibility, ERP-style business tooling, and SaaS productivity platforms. Rather than aggregating hunches, we apply one simple integrity rule: every figure about search engine and AI behavior points to a named, dated study; every price range comes from our own quoting data; and any market figure we cannot source is flagged in plain sight, never invented.

Two anchor figures are enough to set the 2026 scene. The first measures the scale of the transformation of search; the second shows that content quality is no longer optional. Both come from named studies, SearchEngineLand and GenOptima, and they set the frame within which any SMB web strategy must now be designed.

The experiential base of this report is our own field work: ten years in business since 2024, more than seventy clients served and more than fifty projects delivered, across sectors as varied as real estate, hospitality, healthcare and vertical SaaS. As a concrete illustration of that diversity, the portfolio drawn on in this report covers three real, named cases: Servicimmo in real estate, CoProFlex in property management and DocAgora in healthcare. That experience informs the qualitative reading, but it is no substitute for a representative survey. This is a distinction we want to state plainly: Propulseo has not yet run a proprietary survey of a panel of French SMBs. Until that survey is conducted, we publish no equipment rate, no share of SMBs satisfied with their website and no software adoption percentage that would pass itself off as a survey result.

In practice, the report therefore combines two registers. The first gathers verifiable public data on search engine and AI behavior, each figure shown with its source at the point of use. The second consists of market orders of magnitude drawn from our quoting activity, such as the three ranges for websites, ERP and SaaS. Wherever a penetration or perception statistic would be useful but cannot be sourced, we insert an explicit marker rather than an arbitrary number. Line by line, the reader knows what is measured, what is observed and what remains to be quantified.

Three principles guide the reading. First: a figure without a source is worthless, and a survey figure without a survey even more so. Second: digital maturity cannot be reduced to owning a website; it is measured by the ability of that site and of the business tools behind it to produce visibility, leads and productivity. Third: 2026 is not a year of continuity but of tipping points, and an honest report has to say so, numbers in hand. The sections that follow apply this grid to the web, to SEO turned GEO, to ERP, to SaaS and to budgets, before drawing predictions and laying out the limitations.

SMB web maturity in 2026

The web maturity of an SMB is no longer measured by the mere existence of a website. In 2026, the relevant question is whether that site is visible where demand takes shape, and able to turn a visit into a contact. Two structural shifts are redefining the bar. On one side, 60% of Google SERPs display an AI Overview, which pushes the classic blue link further down the page. On the other, conversion depends heavily on page structure: a single-CTA landing page converts at 13.5% versus 10.5% multi-CTA. A professional website can no longer settle for existing: it has to be designed for the generated answer and for action.

In the field, we observe a clear divide between two families of SMBs. The first has a recent site, hand-coded or solidly configured, designed for search rankings and conversion. The second drags along an old, slow, poorly structured site, often built on a pile of plugins that breaks with every update. That second family pays twice: once in lost visibility, once in technical debt to be reworked. Giving a representative measure of how French SMBs split between these two families would require a dedicated survey we have not yet run, and we do not substitute an invented figure for that field observation.

The most underestimated component of web maturity remains content. A technically flawless site that is poor in usable text shows up neither in rich SERPs nor in generated answers. The 2026 data is unequivocal: original data delivers a visibility gain, while paraphrased AI content with no added value collapses. In other words, web maturity is now decided as much in the editorial as in the code.

A concrete example is the Servicimmo project in real estate: the custom website was not conceived as a frozen brochure but as a visibility foundation, with SEO work meant to sustain organic rankings over time. That is the maturity gap between a site that costs and a site that earns. A mature SMB in 2026 does not just ask whether it has a website, but whether that site is a citable, fresh, conversion-oriented source.

From SEO to GEO: the visibility shift

The most structural change of 2026 is not a new algorithm; it is a new place where answers happen. Search is moving from the blue link to the generated answer, and visibility is no longer won in the same places. Two figures sum up the break. First, 60% of Google SERPs display an AI Overview, which diverts part of the clicks. Second, and most revealing, the overlap between Google's top results and the sources actually cited by LLMs falls below 20%. Ranking first on Google therefore no longer guarantees being cited by ChatGPT or Perplexity: these are two distinct competitions.

GEO, or Generative Engine Optimization, answers this new reality. It does not replace SEO; it adds to it, by optimizing the probability that a piece of content will be selected, understood and cited by a generative model. The dominant criterion becomes format. Authoritas data shows that 74.2% of AI citations come from list-structured content, which makes the listicle format a prerequisite, not a style choice. An SMB that wants to exist in generated answers must therefore structure its pages into clear, extractable, hierarchical blocks, where classic SEO tolerated more literary pages.

Freshness is the second GEO lever. With 53% of AI-cited sources less than six months old, content stops being a fixed asset and becomes an asset to maintain. This mechanic is compounded by the sheer scale of generative demand. With more than 800 million weekly ChatGPT users and 780 million monthly queries on Perplexity, a growing share of prospect questions never reaches a classic results page: it gets resolved inside a conversation. Ignoring this channel means accepting progressive invisibility with a massive audience.

Site structure amplifies or throttles these effects. A cluster architecture, with a pillar page linked to satellite pages, gains +40% in rankings according to the Geneo study, because it signals coherent topical coverage to engines and models alike. A concrete example: on Servicimmo, the SEO work did not consist of stacking keywords but of building a readable content structure, the precondition for being picked up both by Google and by generative engines. The SEO-to-GEO shift does not require abandoning what worked; it requires adding a layer of structure, format and freshness.

Business software and ERP adoption

If visibility is the external front of digital, business tooling is the internal one. A mature SMB does not settle for being found; it runs its operations on tools that fit. This is the territory of ERP, or custom business software, with a reference range of EUR 15,000 to 150,000. That envelope does not reflect commercial vagueness but a reality of scope: a first structuring business scope generally fits within EUR 30,000 to 90,000, while a multi-module tool deployed across several departments, with numerous integrations, climbs toward the top of the range.

The way an ERP is priced often surprises executives used to thinking in numbers of screens. On business software, the main cost multiplier is not the interface but integration. Every connection to an accounting tool, payroll software or external API represents a development lot to design, test and maintain. That is why two ERPs with a similar functional scope can vary twofold depending on how interconnected they are. As for precisely measuring the share of French SMBs equipped with an ERP genuinely suited to their business, as opposed to a poorly configured generalist suite, that would require a survey we have not conducted, and we would rather say so than put forward an unverifiable adoption rate.

A concrete example is an ERP project designed for a resort in Thailand. Before the project, operations were coordinated across a mosaic of disparate tools and spreadsheets, a source of delays and errors between field teams and the back office. The answer was a custom ERP centralizing operations management in a single interface, with modules structured around the actual workflows. The case illustrates the maturity gap between an SMB that endures its tools and an SMB that shapes them around its processes.

The trade-off between an off-the-shelf solution and custom development remains central to SMB software adoption. A generalist suite is attractive for its entry price, but it hides three costs: configuration, often lengthy; poor fit, paid for in daily manual workarounds; and cumulative licensing, which runs for as long as the tool is used. Custom makes sense when the business is specific, when processes are a competitive advantage, or when the user count makes licenses heavier than the build. A quantified measure of how these trade-offs split nationally would, again, require a dedicated survey we flag rather than replace with an arbitrary estimate.

Business tooling does not live in isolation from visibility: an ERP that centralizes data also feeds the content that engines and AI systems will cite. On this type of project, consolidating operations in a single interface made publishable information reliable, whereas scattered data produces inconsistent pages. And data quality bears directly on generative visibility: content backed by original data gains +22% visibility, and semantically complete content earns 4.2 times more AI citations. An SMB that structures its business internally therefore gives itself, as a knock-on effect, editorial material more citable than competitors still stuck on manual re-entry.

SaaS and team productivity

SaaS occupies a particular position in SMB digital maturity: it is both a tool you consume and, for some companies, a product you build. On the build side, custom SaaS is the most demanding category, with a range of EUR 40,000 to 300,000. The low end, EUR 40,000 to 80,000, corresponds to an already market-ready MVP with a multi-tenant architecture, subscription billing and a clean technical base. The high end, up to EUR 300,000, targets a mature platform with many roles, integrations and a heavy user load.

At comparable scope, a SaaS costs more than an internal ERP for three structural reasons: data isolation between customers in a multi-tenant architecture, recurring billing, which is a module in its own right, and user autonomy, since users must understand the tool without training. That last point shifts budget toward experience and onboarding, and it ties directly into commercial conversion. A single-CTA entry page converts at 13.5% versus 10.5% multi-CTA, and a CTA tailored to the page context can multiply clicks, which weighs heavily for a product sold online.

Two concrete examples: vertical SaaS products built in-house and for clients. CoProFlex is a property management SaaS that models the workflows specific to co-owned buildings. DocAgora is a custom medical SaaS structured around the constraints of the healthcare sector in Portugal. Both products share a layering logic: start with a controlled scope, then add features as real usage dictates, without rebuilding everything thanks to a sound technical base.

On the consumption side, SMB team productivity is decided by adopting SaaS tools that are well chosen and well integrated. A stack of unconnected subscriptions recreates, only worse, the disparate spreadsheet problem encountered at a hotel resort: scattered data and manual re-entry. Maturity means integrating these tools, or even consolidating some of them into a business SaaS. Quantifying the average number of SaaS tools used by a French SMB, and the share actually integrated with one another, would require a proprietary survey, and until it is run we refrain from advancing a figure.

SMB digital budgets

Three ranges structure the digital budget of an SMB in 2026, each matching a maturity level. A custom website runs from EUR 1,500 to 10,000, including EUR 4,000 to 8,000 for a professional business website, the heart of the market. A custom ERP runs from EUR 15,000 to 150,000. A custom SaaS runs from EUR 40,000 to 300,000. These envelopes are not to be compared with one another: they answer different needs, being visible, running the company, or selling a product.

The most widespread budgeting trap is looking only at the build price. The real cost of a digital project includes hosting, maintenance, third-party licenses, content and evolutions, line items often missing from low-cost quotes. That is the notion of total cost of ownership, projected over three to five years. And content in particular is not a decorative expense: semantically complete content earns 4.2 times more AI citations, and freshness conditions presence in generated answers, since 53% of AI-cited sources are less than six months old. Budgeting without a recurring editorial line means funding an asset that depreciates.

To compare three budget trajectories, the following table sets the three project families side by side: envelope, typical value and the business goal each serves. It helps an SMB locate its own priority before any detailed quote.

The method that secures these budgets is phased delivery. A EUR 90,000 ERP financed in one go strains the cash position of an SMB; the same project delivered module by module lets you use and monetize the first features before funding the next ones. The same logic applies to SaaS, where CoProFlex and DocAgora both followed a progressive layering. Giving a realistic median weight of annual digital budget relative to the revenue of a French SMB would require a survey we have not run, and we leave that magnitude pending rather than put a number on it without a basis.

Finally, two figures help calibrate the effort on conversion and responsiveness, both often neglected in budgets. A trust signal placed near the CTA lifts conversion markedly, and responding to a lead within five minutes multiplies the number of qualified leads. Investing a little budget in these two items often returns more than one extra feature.

Barriers to remove, levers to pull

The digital maturity of an SMB rarely stalls for lack of information: it runs into concrete barriers and moves forward thanks to identifiable levers. On the barrier side, field experience highlights recurring obstacles whose exact weight across the French economy still needs to be measured by survey. The levers, by contrast, can already be quantified from public data, and they are where an SMB has the most to gain by focusing its energy in 2026.

1. The technical debt barrier. A site or a tool built fast and badly is expensive to rework, sometimes as much as a full rebuild. As long as that weight goes unquantified by a dedicated survey, it blocks any investment decision.
2. The quote legibility barrier. A single lump-sum figure prevents any comparison and hides where the money goes. The lever is to demand a phased, itemized breakdown, which makes each line something you can weigh and decide on.
3. The barrier of seeing content as a cost. Many SMBs still treat editorial work as a side expense, when semantically complete content brings 4.2 times more AI citations and original data adds +22% visibility.
4. The commercial responsiveness barrier. A site that generates leads without a fast callback process wastes its investment, when a reply within five minutes multiplies qualified leads a hundredfold.

On the lever side, three high-return actions stand out from the 2026 data. The first is pillar-and-satellite cluster structuring, which brings +40% in rankings and installs durable topical authority. The second is reformatting for extraction, since 74.2% of AI citations come from list-based content: reworking existing pages costs little and unlocks generative visibility. The third is maintaining freshness, as half of the sources cited by AI are less than six months old.

A concrete example is the CoProFlex case: a property management SaaS is not judged by its build cost but by its ability to carry complex administrative, document and financial workflows over time. A cheaper generalist tool would have generated hidden costs of permanent adaptation, a perfect illustration of the false-economy barrier. The right reflex is to relate every expense to its expected return, leads, productivity or recurring revenue, rather than to its sticker price.

2026-2027 predictions

Budget ranges are unlikely to collapse in the short term, but their internal structure is shifting, and the border between web visibility and business tooling keeps blurring. Here are the trends we observe and their consequences for SMBs, anchored in public data wherever possible.

1. GEO becomes a budget line in its own right. With 60% of SERPs showing an AI Overview and less than 20% overlap between Google top results and LLM-cited sources, the share devoted to content structure, format and freshness stops being an extra and becomes central.
2. Freshness becomes an accepted recurring cost. With 53% of AI-cited sources less than six months old, maintaining a site stops being optional; SMBs that budget continuous updates take the lead.
3. Demand migrates to conversational interfaces. ChatGPT's 800 million weekly users and Perplexity's 780 million monthly queries durably install a channel that no longer goes through the classic results page.
4. Phased delivery becomes the norm for ERP and SaaS. Under budget constraints, billing by modules and milestones becomes standard, which makes the EUR 15,000 to 300,000 envelopes more accessible to SMBs.

Concretely, an executive preparing a 2026-2027 budget should no longer think in terms of a one-off website cost, but of an annual envelope covering the amortized build, maintenance and content production, all connected to the company's business tooling. A concrete example: the CoProFlex and DocAgora SaaS products illustrate this progressive layering logic, where investment continues after the initial delivery to follow real usage. Several market magnitudes useful to these predictions, such as the pace of GEO adoption by French SMBs or the evolution of their digital budgets, remain to be quantified by a dedicated survey, and we flag them as such rather than putting numbers on them prematurely.

Detailed methodology and limitations

For this report to be citable without reservation, here is how its figures are built and what they do not say. Data on search engine and AI behavior comes from named, dated studies: SearchEngineLand for the share of AI Overviews; Authoritas for overlap, listicle share and freshness; GenOptima for semantic completeness; SE Ranking for original data; Unbounce and HubSpot for conversion; Directive Consulting for response speed; Geneo for the cluster effect; OpenAI and Perplexity for usage volumes. Each is displayed with its source at the point of use, in the body of the page.

The three price ranges, websites from EUR 1,500 to 10,000, ERP from EUR 15,000 to 150,000 and SaaS from EUR 40,000 to 300,000, are market orders of magnitude drawn from ten years of quoting and more than fifty projects delivered since 2024. The typical values, EUR 4,000 to 8,000 for a business website, EUR 30,000 to 90,000 for a first ERP scope and EUR 40,000 to 80,000 for a SaaS MVP, designate the heart of the distribution, where most SMB projects concentrate, not a floor or a ceiling. A project can fall outside these zones depending on scope, integrations and security requirements.

The most important limitation of this report is owned and repeated: Propulseo has not yet run a proprietary survey of a representative panel of French SMBs. Consequently, no equipment rate, no satisfaction share, no adoption percentage and no median budget is put forward as a survey result. Wherever such a figure would be relevant, we placed an explicit marker rather than an invented number. The proprietary data points still to be produced concern, among others, the share of SMBs with a high-performing website, the adoption rate of business-fit ERP, the off-the-shelf versus custom split, the number and integration of SaaS tools, the weight of technical debt, the median digital budget and the pace of GEO adoption. This set of indicators remains to be quantified by the proprietary survey we plan to run.

No quantified metric has been attributed to the client cases cited, CoProFlex, DocAgora and Servicimmo: until a result is confirmed by the client, it stays qualitative, as a matter of integrity. The most reliable way to turn the observations on this page into quantified decisions remains a framing diagnostic, which relates each stake, visibility, productivity or revenue, to a scope and to a budget defensible lot by lot. This report is meant to be updated as soon as the proprietary survey has been conducted, replacing the markers with sourced figures.